1 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-32 2 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple i386-pc-linux-gnu -emit-pch -o %t %s 3 // RUN: %clang_cc1 -fopenmp -x c++ -triple i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-32 4 // RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DLAMBDA -triple i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck -check-prefix=LAMBDA -check-prefix=LAMBDA-32 %s 5 // RUN: %clang_cc1 -verify -fopenmp -x c++ -fblocks -DBLOCKS -triple i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck -check-prefix=BLOCKS -check-prefix=BLOCKS-32 %s 6 7 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s 8 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple i386-pc-linux-gnu -emit-pch -o %t %s 9 // RUN: %clang_cc1 -fopenmp-simd -x c++ -triple i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY0 %s 10 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -std=c++11 -DLAMBDA -triple i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s 11 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -fblocks -DBLOCKS -triple i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s 12 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}} 13 14 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-64 15 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-pc-linux-gnu -emit-pch -o %t %s 16 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-64 17 // RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DLAMBDA -triple x86_64-pc-linux-gnu -emit-llvm %s -o - | FileCheck -check-prefix=LAMBDA -check-prefix=LAMBDA-64 %s 18 // RUN: %clang_cc1 -verify -fopenmp -x c++ -fblocks -DBLOCKS -triple x86_64-pc-linux-gnu -emit-llvm %s -o - | FileCheck -check-prefix=BLOCKS -check-prefix=BLOCKS-64 %s 19 20 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple x86_64-pc-linux-gnu -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY1 %s 21 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-pc-linux-gnu -emit-pch -o %t %s 22 // RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY1 %s 23 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -std=c++11 -DLAMBDA -triple x86_64-pc-linux-gnu -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY1 %s 24 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -fblocks -DBLOCKS -triple x86_64-pc-linux-gnu -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY1 %s 25 // SIMD-ONLY1-NOT: {{__kmpc|__tgt}} 26 27 // 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 28 29 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -std=c++11 -DARRAY -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY2 %s 30 // SIMD-ONLY2-NOT: {{__kmpc|__tgt}} 31 // expected-no-diagnostics 32 #ifndef ARRAY 33 #ifndef HEADER 34 #define HEADER 35 36 enum omp_allocator_handle_t { 37 omp_null_allocator = 0, 38 omp_default_mem_alloc = 1, 39 omp_large_cap_mem_alloc = 2, 40 omp_const_mem_alloc = 3, 41 omp_high_bw_mem_alloc = 4, 42 omp_low_lat_mem_alloc = 5, 43 omp_cgroup_mem_alloc = 6, 44 omp_pteam_mem_alloc = 7, 45 omp_thread_mem_alloc = 8, 46 KMP_ALLOCATOR_MAX_HANDLE = __UINTPTR_MAX__ 47 }; 48 49 struct St { 50 int a, b; 51 St() : a(0), b(0) {} 52 St(const St &st) : a(st.a + st.b), b(0) {} 53 ~St() {} 54 }; 55 56 volatile int g __attribute__((aligned(128))) = 1212; 57 58 struct SS { 59 int a; 60 int b : 4; 61 int &c; 62 int e[4]; 63 SS(int &d) : a(0), b(0), c(d) { 64 #pragma omp parallel firstprivate(a, b, c, e) 65 #ifdef LAMBDA 66 [&]() { 67 ++this->a, --b, (this)->c /= 1; 68 #pragma omp parallel firstprivate(a, b, c) 69 ++(this)->a, --b, this->c /= 1; 70 }(); 71 #elif defined(BLOCKS) 72 ^{ 73 ++a; 74 --this->b; 75 (this)->c /= 1; 76 #pragma omp parallel firstprivate(a, b, c) 77 ++(this)->a, --b, this->c /= 1; 78 }(); 79 #else 80 ++this->a, --b, c /= 1, e[2] = 1111; 81 #endif 82 } 83 }; 84 85 template<typename T> 86 struct SST { 87 T a; 88 SST() : a(T()) { 89 #pragma omp parallel firstprivate(a) 90 #ifdef LAMBDA 91 [&]() { 92 [&]() { 93 ++this->a; 94 #pragma omp parallel firstprivate(a) 95 ++(this)->a; 96 }(); 97 }(); 98 #elif defined(BLOCKS) 99 ^{ 100 ^{ 101 ++a; 102 #pragma omp parallel firstprivate(a) 103 ++(this)->a; 104 }(); 105 }(); 106 #else 107 ++(this)->a; 108 #endif 109 } 110 }; 111 112 template <class T> 113 struct S { 114 T f; 115 S(T a) : f(a + g) {} 116 S() : f(g) {} 117 S(const S &s, St t = St()) : f(s.f + t.a) {} 118 operator T() { return T(); } 119 ~S() {} 120 }; 121 122 // CHECK: [[SS_TY:%.+]] = type { i{{[0-9]+}}, i8 123 // LAMBDA: [[SS_TY:%.+]] = type { i{{[0-9]+}}, i8 124 // BLOCKS: [[SS_TY:%.+]] = type { i{{[0-9]+}}, i8 125 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { float } 126 // CHECK-DAG: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} } 127 // CHECK-DAG: [[ST_TY:%.+]] = type { i{{[0-9]+}}, i{{[0-9]+}} } 128 129 template <typename T> 130 T tmain() { 131 S<T> test; 132 SST<T> sst; 133 T t_var __attribute__((aligned(128))) = T(); 134 T vec[] __attribute__((aligned(128))) = {1, 2}; 135 S<T> s_arr[] __attribute__((aligned(128))) = {1, 2}; 136 S<T> var __attribute__((aligned(128))) (3); 137 #pragma omp parallel firstprivate(t_var, vec, s_arr, var) 138 { 139 vec[0] = t_var; 140 s_arr[0] = var; 141 } 142 #pragma omp parallel firstprivate(t_var) 143 {} 144 return T(); 145 } 146 147 int main() { 148 static int sivar; 149 SS ss(sivar); 150 #ifdef LAMBDA 151 // LAMBDA: [[G:@.+]] ={{.*}} global i{{[0-9]+}} 1212, 152 // LAMBDA-LABEL: @main 153 // LAMBDA: alloca [[SS_TY]], 154 // LAMBDA: alloca [[CAP_TY:%.+]], 155 // LAMBDA: call{{.*}} void [[OUTER_LAMBDA:@[^(]+]]([[CAP_TY]]* 156 [&]() { 157 // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]]( 158 // LAMBDA: call {{.*}}void {{.+}} @__kmpc_fork_call({{.+}}, i32 2, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}, i32* [[G]], {{.+}}) 159 #pragma omp parallel firstprivate(g, sivar) 160 { 161 // LAMBDA: define {{.+}} @{{.+}}([[SS_TY]]* 162 // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0 163 // LAMBDA: store i{{[0-9]+}} 0, i{{[0-9]+}}* % 164 // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1 165 // LAMBDA: store i8 166 // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2 167 // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0 168 // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1 169 // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2 170 // 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]]*, [[iz:i64|i32]], {{i64|i32}}, {{i64|i32}}, [4 x i{{[0-9]+}}]*)* [[SS_MICROTASK:@.+]] to void 171 // LAMBDA: ret 172 173 // LAMBDA: define internal void [[SS_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[SS_TY]]* %{{.+}}, [[iz]] {{.+}}, [[iz]] {{.+}}, [[iz]] {{.+}}, [4 x i{{[0-9]+}}]* {{.+}}) 174 // LAMBDA-NOT: getelementptr {{.*}}[[SS_TY]], [[SS_TY]]* % 175 // LAMBDA: call{{.*}} void 176 // LAMBDA: ret void 177 178 // LAMBDA: define internal void @{{.+}}(i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[SS_TY]]* %{{.+}}, [[iz]] {{.+}}, [[iz]] {{.+}}, [[iz]] {{.+}}) 179 // LAMBDA: [[A_PRIV:%.+]] = alloca i{{[0-9]+}}, 180 // LAMBDA: [[B_PRIV:%.+]] = alloca i{{[0-9]+}}, 181 // LAMBDA: [[C_PRIV:%.+]] = alloca i{{[0-9]+}}, 182 // LAMBDA-64: [[A_CONV:%.+]] = bitcast i64* [[A_PRIV]] to i32* 183 // LAMBDA-64: [[B_CONV:%.+]] = bitcast i64* [[B_PRIV]] to i32* 184 // LAMBDA-64: [[C_CONV:%.+]] = bitcast i64* [[C_PRIV]] to i32* 185 // LAMBDA-64: store i32* [[A_CONV]], i32** [[REFA:%.+]], 186 // LAMBDA-32: store i32* [[A_PRIV]], i32** [[REFA:%.+]], 187 // LAMBDA-64: store i32* [[C_CONV]], i32** [[REFC:%.+]], 188 // LAMBDA-32: store i32* [[C_PRIV]], i32** [[REFC:%.+]], 189 // LAMBDA-NEXT: [[A_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFA]], 190 // LAMBDA-NEXT: [[A_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[A_PRIV]], 191 // LAMBDA-NEXT: [[INC:%.+]] = add nsw i{{[0-9]+}} [[A_VAL]], 1 192 // LAMBDA-NEXT: store i{{[0-9]+}} [[INC]], i{{[0-9]+}}* [[A_PRIV]], 193 // LAMBDA-64-NEXT: [[B_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[B_CONV]], 194 // LAMBDA-32-NEXT: [[B_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[B_PRIV]], 195 // LAMBDA-NEXT: [[DEC:%.+]] = add nsw i{{[0-9]+}} [[B_VAL]], -1 196 // LAMBDA-64-NEXT: store i{{[0-9]+}} [[DEC]], i{{[0-9]+}}* [[B_CONV]], 197 // LAMBDA-32-NEXT: store i{{[0-9]+}} [[DEC]], i{{[0-9]+}}* [[B_PRIV]], 198 // LAMBDA-NEXT: [[C_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFC]], 199 // LAMBDA-NEXT: [[C_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[C_PRIV]], 200 // LAMBDA-NEXT: [[DIV:%.+]] = sdiv i{{[0-9]+}} [[C_VAL]], 1 201 // LAMBDA-NEXT: store i{{[0-9]+}} [[DIV]], i{{[0-9]+}}* [[C_PRIV]], 202 // LAMBDA-NEXT: ret void 203 204 // LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}, i32* nonnull align 4 dereferenceable(4) %{{.+}}, [[iz]] {{.*}}%{{.+}}) 205 // LAMBDA: [[SIVAR_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}}, 206 // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}}, align 128 207 // LAMBDA: [[G_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[G_REF_ADDR:%.+]] 208 // LAMBDA-64: [[SIVAR_PRIVATE_CONV:%.+]] = bitcast i64* [[SIVAR_PRIVATE_ADDR]] to i32* 209 // LAMBDA: [[G_VAL:%.+]] = load volatile i{{[0-9]+}}, i{{[0-9]+}}* [[G_REF]], align 128 210 // LAMBDA: store i{{[0-9]+}} [[G_VAL]], i{{[0-9]+}}* [[G_PRIVATE_ADDR]], align 128 211 // LAMBDA-NOT: call {{.*}}void @__kmpc_barrier( 212 g = 1; 213 sivar = 2; 214 // LAMBDA: store i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]], 215 // LAMBDA-64: store i{{[0-9]+}} 2, i{{[0-9]+}}* [[SIVAR_PRIVATE_CONV]], 216 // LAMBDA-32: store i{{[0-9]+}} 2, i{{[0-9]+}}* [[SIVAR_PRIVATE_ADDR]], 217 // LAMBDA: [[G_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 218 // LAMBDA: store i{{[0-9]+}}* [[G_PRIVATE_ADDR]], i{{[0-9]+}}** [[G_PRIVATE_ADDR_REF]] 219 // LAMBDA: [[SIVAR_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 1 220 // LAMBDA-64: store i{{[0-9]+}}* [[SIVAR_PRIVATE_CONV]], i{{[0-9]+}}** [[SIVAR_PRIVATE_ADDR_REF]] 221 // LAMBDA-32: store i{{[0-9]+}}* [[SIVAR_PRIVATE_ADDR]], i{{[0-9]+}}** [[SIVAR_PRIVATE_ADDR_REF]] 222 // LAMBDA: call{{.*}} void [[INNER_LAMBDA:@.+]](%{{.+}}* {{[^,]*}} [[ARG]]) 223 [&]() { 224 // LAMBDA: define {{.+}} void [[INNER_LAMBDA]](%{{.+}}* {{[^,]*}} [[ARG_PTR:%.+]]) 225 // LAMBDA: store %{{.+}}* [[ARG_PTR]], %{{.+}}** [[ARG_PTR_REF:%.+]], 226 g = 2; 227 sivar = 4; 228 // LAMBDA: [[ARG_PTR:%.+]] = load %{{.+}}*, %{{.+}}** [[ARG_PTR_REF]] 229 // LAMBDA: [[G_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 230 // LAMBDA: [[G_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[G_PTR_REF]] 231 // LAMBDA: [[SIVAR_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 1 232 // LAMBDA: [[SIVAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[SIVAR_PTR_REF]] 233 // LAMBDA: store i{{[0-9]+}} 4, i{{[0-9]+}}* [[SIVAR_REF]] 234 }(); 235 } 236 }(); 237 return 0; 238 #elif defined(BLOCKS) 239 // BLOCKS: [[G:@.+]] ={{.*}} global i{{[0-9]+}} 1212, 240 // BLOCKS: [[SIVAR:@.+]] = internal global i{{[0-9]+}} 0, 241 // BLOCKS-LABEL: @main 242 // BLOCKS: call 243 // BLOCKS: call {{.*}}void {{%.+}}(i8 244 ^{ 245 // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8* 246 // BLOCKS: call {{.*}}void {{.+}} @__kmpc_fork_call({{.+}}, i32 2, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}, i32* [[G]], {{.+}}) 247 #pragma omp parallel firstprivate(g, sivar) 248 { 249 // BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}, i32* nonnull align 4 dereferenceable(4) %{{.+}}, [[iz:i64|i32]] {{.*}}%{{.+}}) 250 // BLOCKS: [[SIVAR_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}}, 251 // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}}, align 128 252 // BLOCKS: [[G_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[G_REF_ADDR:%.+]] 253 // BLOCKS-64: [[SIVAR_PRIVATE_CONV:%.+]] = bitcast i64* [[SIVAR_PRIVATE_ADDR]] to i32* 254 // BLOCKS: [[G_VAL:%.+]] = load volatile i{{[0-9]+}}, i{{[0-9]+}}* [[G_REF]], align 128 255 // BLOCKS: store i{{[0-9]+}} [[G_VAL]], i{{[0-9]+}}* [[G_PRIVATE_ADDR]], align 128 256 // BLOCKS-NOT: call {{.*}}void @__kmpc_barrier( 257 g = 1; 258 sivar = 2; 259 // BLOCKS: store i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]], 260 // BLOCKS-64: store i{{[0-9]+}} 2, i{{[0-9]+}}* [[SIVAR_PRIVATE_CONV]], 261 // BLOCKS-32: store i{{[0-9]+}} 2, i{{[0-9]+}}* [[SIVAR_PRIVATE_ADDR]], 262 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 263 // BLOCKS: i{{[0-9]+}}* [[G_PRIVATE_ADDR]] 264 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 265 // BLOCKS-NOT: [[SIVAR]]{{[[^:word:]]}} 266 // BLOCKS-64: i{{[0-9]+}}* [[SIVAR_PRIVATE_CONV]] 267 // BLOCKS-32: i{{[0-9]+}}* [[SIVAR_PRIVATE_ADDR]] 268 // BLOCKS-NOT: [[SIVAR]]{{[[^:word:]]}} 269 // BLOCKS: call {{.*}}void {{%.+}}(i8 270 ^{ 271 // BLOCKS: define {{.+}} void {{@.+}}(i8* 272 g = 2; 273 sivar = 4; 274 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 275 // BLOCKS: store i{{[0-9]+}} 2, i{{[0-9]+}}* 276 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 277 // BLOCKS-NOT: [[SIVAR]]{{[[^:word:]]}} 278 // BLOCKS: store i{{[0-9]+}} 4, i{{[0-9]+}}* 279 // BLOCKS-NOT: [[SIVAR]]{{[[^:word:]]}} 280 // BLOCKS: ret 281 }(); 282 } 283 }(); 284 return 0; 285 // BLOCKS: define {{.+}} @{{.+}}([[SS_TY]]* 286 // BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0 287 // BLOCKS: store i{{[0-9]+}} 0, i{{[0-9]+}}* % 288 // BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1 289 // BLOCKS: store i8 290 // BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2 291 // BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0 292 // BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1 293 // BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2 294 // 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]]*, [[iz]], [[iz]], [[iz]], [4 x i{{[0-9]+}}]*)* [[SS_MICROTASK:@.+]] to void 295 // BLOCKS: ret 296 297 // BLOCKS: define internal void [[SS_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[SS_TY]]* %{{.+}}, [[iz]] {{.+}}, [[iz]] {{.+}}, [[iz]] {{.+}}, [4 x i{{[0-9]+}}]* {{.+}}) 298 // BLOCKS-NOT: getelementptr {{.*}}[[SS_TY]], [[SS_TY]]* % 299 // BLOCKS: call{{.*}} void 300 // BLOCKS: ret void 301 302 // BLOCKS: define internal void @{{.+}}(i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[SS_TY]]* %{{.+}}, [[iz]] {{.+}}, [[iz]] {{.+}}, [[iz]] {{.+}}) 303 // BLOCKS: [[A_PRIV:%.+]] = alloca i{{[0-9]+}}, 304 // BLOCKS: [[B_PRIV:%.+]] = alloca i{{[0-9]+}}, 305 // BLOCKS: [[C_PRIV:%.+]] = alloca i{{[0-9]+}}, 306 // BLOCKS-64: [[A_CONV:%.+]] = bitcast i64* [[A_PRIV]] to i32* 307 // BLOCKS-64: [[B_CONV:%.+]] = bitcast i64* [[B_PRIV]] to i32* 308 // BLOCKS-64: [[C_CONV:%.+]] = bitcast i64* [[C_PRIV]] to i32* 309 // BLOCKS-64: store i32* [[A_CONV]], i32** [[REFA:%.+]], 310 // BLOCKS-32: store i32* [[A_PRIV]], i32** [[REFA:%.+]], 311 // BLOCKS-64: store i32* [[C_CONV]], i32** [[REFC:%.+]], 312 // BLOCKS-32: store i32* [[C_PRIV]], i32** [[REFC:%.+]], 313 // BLOCKS-NEXT: [[A_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFA]], 314 // BLOCKS-NEXT: [[A_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[A_PRIV]], 315 // BLOCKS-NEXT: [[INC:%.+]] = add nsw i{{[0-9]+}} [[A_VAL]], 1 316 // BLOCKS-NEXT: store i{{[0-9]+}} [[INC]], i{{[0-9]+}}* [[A_PRIV]], 317 // BLOCKS-64-NEXT: [[B_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[B_CONV]], 318 // BLOCKS-32-NEXT: [[B_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[B_PRIV]], 319 // BLOCKS-NEXT: [[DEC:%.+]] = add nsw i{{[0-9]+}} [[B_VAL]], -1 320 // BLOCKS-64-NEXT: store i{{[0-9]+}} [[DEC]], i{{[0-9]+}}* [[B_CONV]], 321 // BLOCKS-32-NEXT: store i{{[0-9]+}} [[DEC]], i{{[0-9]+}}* [[B_PRIV]], 322 // BLOCKS-NEXT: [[C_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFC]], 323 // BLOCKS-NEXT: [[C_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[C_PRIV]], 324 // BLOCKS-NEXT: [[DIV:%.+]] = sdiv i{{[0-9]+}} [[C_VAL]], 1 325 // BLOCKS-NEXT: store i{{[0-9]+}} [[DIV]], i{{[0-9]+}}* [[C_PRIV]], 326 // BLOCKS-NEXT: ret void 327 #else 328 S<float> test; 329 int t_var = 0; 330 int vec[] = {1, 2}; 331 S<float> s_arr[] = {1, 2}; 332 S<float> var(3); 333 #pragma omp parallel firstprivate(t_var, vec, s_arr, var, sivar) 334 { 335 vec[0] = t_var; 336 s_arr[0] = var; 337 sivar = 2; 338 } 339 const int a = 0; 340 #pragma omp parallel allocate(omp_default_mem_alloc: t_var) firstprivate(t_var, a) 341 { t_var = a; } 342 return tmain<int>(); 343 #endif 344 } 345 346 // CHECK: define {{.*}}i{{[0-9]+}} @main() 347 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]], 348 // CHECK: [[T_VAR:%.+]] = alloca i32, 349 // CHECK: [[T_VARCAST:%.+]] = alloca [[iz:i64|i32]], 350 // CHECK: [[SIVARCAST:%.+]] = alloca [[iz]], 351 // CHECK: [[A:%.+]] = alloca i32, 352 // CHECK: [[T_VARCAST1:%.+]] = alloca [[iz:i64|i32]], 353 // CHECK: call {{.*}} [[S_FLOAT_TY_DEF_CONSTR:@.+]]([[S_FLOAT_TY]]* {{[^,]*}} [[TEST]]) 354 // CHECK: [[T_VARVAL:%.+]] = load i32, i32* [[T_VAR]], 355 // CHECK-64: [[T_VARCONV:%.+]] = bitcast i64* [[T_VARCAST]] to i32* 356 // CHECK-64: store i32 [[T_VARVAL]], i32* [[T_VARCONV]], 357 // CHECK-32: store i32 [[T_VARVAL]], i32* [[T_VARCAST]], 358 // CHECK: [[T_VARPVT:%.+]] = load [[iz]], [[iz]]* [[T_VARCAST]], 359 // CHECK: [[SIVARVAL:%.+]] = load i32, i32* @{{.+}}, 360 // CHECK-64: [[SIVARCONV:%.+]] = bitcast i64* [[SIVARCAST]] to i32* 361 // CHECK-64: store i32 [[SIVARVAL]], i32* [[SIVARCONV]], 362 // CHECK-32: store i32 [[SIVARVAL]], i32* [[SIVARCAST]], 363 // CHECK: [[SIVARPVT:%.+]] = load [[iz]], [[iz]]* [[SIVARCAST]], 364 // 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]*, [[iz]], [2 x [[S_FLOAT_TY]]]*, [[S_FLOAT_TY]]*, i{{[0-9]+}})* [[MAIN_MICROTASK:@.+]] to void {{.*}}[[iz]] [[T_VARPVT]],{{.*}}[[iz]] [[SIVARPVT]] 365 // CHECK: [[T_VARVAL:%.+]] = load i32, i32* [[T_VAR]], 366 // CHECK-64: [[T_VARCONV:%.+]] = bitcast i64* [[T_VARCAST1]] to i32* 367 // CHECK-64: store i32 [[T_VARVAL]], i32* [[T_VARCONV]], 368 // CHECK-32: store i32 [[T_VARVAL]], i32* [[T_VARCAST1]], 369 // CHECK: [[T_VARPVT:%.+]] = load [[iz]], [[iz]]* [[T_VARCAST1]], 370 // CHECK: call {{.*}}void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 1, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [[iz]])* [[MAIN_MICROTASK1:@.+]] to void {{.*}}[[iz]] [[T_VARPVT]]) 371 // CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT:@.+]]() 372 // CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]* 373 // CHECK: ret 374 // 375 // CHECK: define internal {{.*}}void [[MAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [2 x i32]* nonnull align 4 dereferenceable(8) %{{.+}}, [[iz]] {{.*}}%{{.+}}, [2 x [[S_FLOAT_TY]]]* nonnull align 4 dereferenceable(8) %{{.+}}, [[S_FLOAT_TY]]* nonnull align 4 dereferenceable(4) %{{.+}}, [[iz]] {{.*}}[[SIVAR:%.+]]) 376 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 377 // CHECK: [[SIVAR7_PRIV:%.+]] = alloca i{{[0-9]+}}, 378 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}], 379 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]], 380 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 381 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 382 383 // CHECK: [[VEC_REF:%.+]] = load [2 x i{{[0-9]+}}]*, [2 x i{{[0-9]+}}]** % 384 // CHECK-NOT: load i{{[0-9]+}}*, i{{[0-9]+}}** % 385 // CHECK-64: [[T_VAR_CONV:%.+]] = bitcast i64* [[T_VAR_PRIV]] to i32* 386 // CHECK: [[S_ARR_REF:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** % 387 // CHECK: [[VAR_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** % 388 // CHECK-NOT: load i{{[0-9]+}}*, i{{[0-9]+}}** % 389 // CHECK-64: [[SIVAR7_CONV:%.+]] = bitcast i64* [[SIVAR7_PRIV]] to i32* 390 // CHECK: [[VEC_DEST:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_PRIV]] to i8* 391 // CHECK: [[VEC_SRC:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_REF]] to i8* 392 // CHECK: call void @llvm.memcpy.{{.+}}(i8* align {{[0-9]+}} [[VEC_DEST]], i8* align {{[0-9]+}} [[VEC_SRC]], 393 // 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]+}} 0 394 // CHECK: [[S_ARR_BEGIN:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[S_ARR_REF]] to [[S_FLOAT_TY]]* 395 // CHECK: [[S_ARR_PRIV_END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[S_ARR_PRIV_BEGIN]], i{{[0-9]+}} 2 396 // CHECK: [[IS_EMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[S_ARR_PRIV_BEGIN]], [[S_ARR_PRIV_END]] 397 // CHECK: br i1 [[IS_EMPTY]], label %[[S_ARR_BODY_DONE:.+]], label %[[S_ARR_BODY:.+]] 398 // CHECK: [[S_ARR_BODY]] 399 // CHECK: call {{.*}} [[ST_TY_DEFAULT_CONSTR:@.+]]([[ST_TY]]* {{[^,]*}} [[ST_TY_TEMP:%.+]]) 400 // CHECK: call {{.*}} [[S_FLOAT_TY_COPY_CONSTR:@.+]]([[S_FLOAT_TY]]* {{.+}}, [[S_FLOAT_TY]]* {{.+}}, [[ST_TY]]* [[ST_TY_TEMP]]) 401 // CHECK: call {{.*}} [[ST_TY_DESTR:@.+]]([[ST_TY]]* {{[^,]*}} [[ST_TY_TEMP]]) 402 // CHECK: br i1 {{.+}}, label %{{.+}}, label %[[S_ARR_BODY]] 403 // CHECK: call {{.*}} [[ST_TY_DEFAULT_CONSTR]]([[ST_TY]]* {{[^,]*}} [[ST_TY_TEMP:%.+]]) 404 // CHECK: call {{.*}} [[S_FLOAT_TY_COPY_CONSTR]]([[S_FLOAT_TY]]* {{[^,]*}} [[VAR_PRIV]], [[S_FLOAT_TY]]* {{.*}} [[VAR_REF]], [[ST_TY]]* [[ST_TY_TEMP]]) 405 // CHECK: call {{.*}} [[ST_TY_DESTR]]([[ST_TY]]* {{[^,]*}} [[ST_TY_TEMP]]) 406 407 // CHECK-64: store i{{[0-9]+}} 2, i{{[0-9]+}}* [[SIVAR7_CONV]], 408 // CHECK-32: store i{{[0-9]+}} 2, i{{[0-9]+}}* [[SIVAR7_PRIV]], 409 410 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* {{[^,]*}} [[VAR_PRIV]]) 411 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* 412 // CHECK: ret void 413 414 415 // CHECK: define internal void [[MAIN_MICROTASK1]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[iz]] [[T_VAR:%.+]]) 416 // CHECK: [[GTID_ADDR:%.+]] = alloca i32*, 417 // CHECK: store [[iz]] [[T_VAR]], [[iz]]* [[T_VAR_ADDR:%.+]], 418 // CHECK-64: [[BC:%.+]] = bitcast [[iz]]* [[T_VAR_ADDR]] to i32* 419 // CHECK: [[GTID_PTR:%.+]] = load i32*, i32** [[GTID_ADDR]], 420 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_PTR]], 421 // CHECK: [[T_VAR_VOID_PTR:%.+]] = call i8* @__kmpc_alloc(i32 [[GTID]], [[iz]] 4, i8* inttoptr ([[iz]] 1 to i8*)) 422 // CHECK: [[T_VAR_PRIV:%.+]] = bitcast i8* [[T_VAR_VOID_PTR]] to i32* 423 // CHECK-32: [[T_VAR_VAL:%.+]] = load i32, i32* [[T_VAR_ADDR]], 424 // CHECK-64: [[T_VAR_VAL:%.+]] = load i32, i32* [[BC]], 425 // CHECK: store i32 [[T_VAR_VAL]], i32* [[T_VAR_PRIV]], 426 // CHECK: store i32 0, i32* [[T_VAR_PRIV]], 427 // CHECK: [[T_VAR_VOID_PTR:%.+]] = bitcast i32* [[T_VAR_PRIV]] to i8* 428 // CHECK: call void @__kmpc_free(i32 [[GTID]], i8* [[T_VAR_VOID_PTR]], i8* inttoptr ([[iz]] 1 to i8*)) 429 // CHECK: ret void 430 431 432 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT]]() 433 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]], 434 // CHECK: call {{.*}} [[S_INT_TY_DEF_CONSTR:@.+]]([[S_INT_TY]]* {{[^,]*}} [[TEST]]) 435 // 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 436 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]* 437 // CHECK: ret 438 // 439 // CHECK: define {{.+}} @{{.+}}([[SS_TY]]* 440 // CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0 441 // CHECK: store i{{[0-9]+}} 0, i{{[0-9]+}}* % 442 // CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1 443 // CHECK: store i8 444 // CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2 445 // CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0 446 // CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1 447 // CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2 448 // 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]]*, [[iz]], [[iz]], [[iz]], [4 x i32]*)* [[SS_MICROTASK:@.+]] to void 449 // CHECK: ret 450 451 // CHECK: define internal void [[SS_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[SS_TY]]* %{{.+}}, [[iz]] {{.+}}, [[iz]] {{.+}}, [[iz]] {{.+}}, [4 x i{{[0-9]+}}]* {{.+}}) 452 // CHECK: [[A_PRIV:%.+]] = alloca i{{[0-9]+}}, 453 // CHECK: [[B_PRIV:%.+]] = alloca i{{[0-9]+}}, 454 // CHECK: [[C_PRIV:%.+]] = alloca i{{[0-9]+}}, 455 // CHECK: [[E_PRIV:%.+]] = alloca [4 x i{{[0-9]+}}], 456 // CHECK: store i{{[0-9]+}} {{.+}}, i{{[0-9]+}}* [[A_PRIV]] 457 // CHECK: store i{{[0-9]+}} {{.+}}, i{{[0-9]+}}* [[B_PRIV]] 458 // CHECK: store i{{[0-9]+}} {{.+}}, i{{[0-9]+}}* [[C_PRIV]] 459 // CHECK-64: [[A_CONV:%.+]] = bitcast i64* [[A_PRIV:%.+]] to i32* 460 // CHECK-64: [[B_CONV:%.+]] = bitcast i64* [[B_PRIV:%.+]] to i32* 461 // CHECK-64: [[C_CONV:%.+]] = bitcast i64* [[C_PRIV:%.+]] to i32* 462 // CHECK-64: store i32* [[A_CONV]], i32** [[REFA:%.+]], 463 // CHECK-32: store i32* [[A_PRIV]], i32** [[REFA:%.+]], 464 // CHECK-64: store i32* [[C_CONV]], i32** [[REFC:%.+]], 465 // CHECK-32: store i32* [[C_PRIV]], i32** [[REFC:%.+]], 466 // CHECK: bitcast [4 x i{{[0-9]+}}]* [[E_PRIV]] to i8* 467 // CHECK: bitcast [4 x i{{[0-9]+}}]* %{{.+}} to i8* 468 // CHECK: call void @llvm.memcpy 469 // CHECK: store [4 x i{{[0-9]+}}]* [[E_PRIV]], [4 x i{{[0-9]+}}]** [[REFE:%.+]], 470 // CHECK-NEXT: [[A_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFA]], 471 // CHECK-NEXT: [[A_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[A_PRIV]], 472 // CHECK-NEXT: [[INC:%.+]] = add nsw i{{[0-9]+}} [[A_VAL]], 1 473 // CHECK-NEXT: store i{{[0-9]+}} [[INC]], i{{[0-9]+}}* [[A_PRIV]], 474 // CHECK-64-NEXT: [[B_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[B_CONV]], 475 // CHECK-32-NEXT: [[B_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[B_PRIV]], 476 // CHECK-NEXT: [[DEC:%.+]] = add nsw i{{[0-9]+}} [[B_VAL]], -1 477 // CHECK-64-NEXT: store i{{[0-9]+}} [[DEC]], i{{[0-9]+}}* [[B_CONV]], 478 // CHECK-32-NEXT: store i{{[0-9]+}} [[DEC]], i{{[0-9]+}}* [[B_PRIV]], 479 // CHECK-NEXT: [[C_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFC]], 480 // CHECK-NEXT: [[C_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[C_PRIV]], 481 // CHECK-NEXT: [[DIV:%.+]] = sdiv i{{[0-9]+}} [[C_VAL]], 1 482 // CHECK-NEXT: store i{{[0-9]+}} [[DIV]], i{{[0-9]+}}* [[C_PRIV]], 483 // CHECK-NEXT: [[E_PRIV:%.+]] = load [4 x i{{[0-9]+}}]*, [4 x i{{[0-9]+}}]** [[REFE]], 484 // 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]+}} 2 485 // CHECK-NEXT: store i32 1111, i32* [[E_PRIV_2]], 486 // CHECK-NEXT: ret void 487 488 // CHECK: define internal {{.*}}void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [2 x i32]* nonnull align 4 dereferenceable(8) %{{.+}}, i32* nonnull align 4 dereferenceable(4) %{{.+}}, [2 x [[S_INT_TY]]]* nonnull align 4 dereferenceable(8) %{{.+}}, [[S_INT_TY]]* nonnull align 4 dereferenceable(4) %{{.+}}) 489 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, align 128 490 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}], align 128 491 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]], align 128 492 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], align 128 493 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 494 495 // CHECK: [[VEC_REF:%.+]] = load [2 x i{{[0-9]+}}]*, [2 x i{{[0-9]+}}]** % 496 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 497 // CHECK: [[S_ARR_REF:%.+]] = load [2 x [[S_INT_TY]]]*, [2 x [[S_INT_TY]]]** % 498 // CHECK: [[VAR_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 499 500 // CHECK: [[T_VAR_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_REF]], align 128 501 // CHECK: store i{{[0-9]+}} [[T_VAR_VAL]], i{{[0-9]+}}* [[T_VAR_PRIV]], align 128 502 // CHECK: [[VEC_DEST:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_PRIV]] to i8* 503 // CHECK: [[VEC_SRC:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_REF]] to i8* 504 // CHECK: call void @llvm.memcpy.{{.+}}(i8* align 128 [[VEC_DEST]], i8* align 128 [[VEC_SRC]], i{{[0-9]+}} {{[0-9]+}}, i1 505 // 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]+}} 0 506 // CHECK: [[S_ARR_BEGIN:%.+]] = bitcast [2 x [[S_INT_TY]]]* [[S_ARR_REF]] to [[S_INT_TY]]* 507 // CHECK: [[S_ARR_PRIV_END:%.+]] = getelementptr [[S_INT_TY]], [[S_INT_TY]]* [[S_ARR_PRIV_BEGIN]], i{{[0-9]+}} 2 508 // CHECK: [[IS_EMPTY:%.+]] = icmp eq [[S_INT_TY]]* [[S_ARR_PRIV_BEGIN]], [[S_ARR_PRIV_END]] 509 // CHECK: br i1 [[IS_EMPTY]], label %[[S_ARR_BODY_DONE:.+]], label %[[S_ARR_BODY:.+]] 510 // CHECK: [[S_ARR_BODY]] 511 // CHECK: call {{.*}} [[ST_TY_DEFAULT_CONSTR]]([[ST_TY]]* {{[^,]*}} [[ST_TY_TEMP:%.+]]) 512 // CHECK: call {{.*}} [[S_INT_TY_COPY_CONSTR:@.+]]([[S_INT_TY]]* {{.+}}, [[S_INT_TY]]* {{.+}}, [[ST_TY]]* [[ST_TY_TEMP]]) 513 // CHECK: call {{.*}} [[ST_TY_DESTR]]([[ST_TY]]* {{[^,]*}} [[ST_TY_TEMP]]) 514 // CHECK: br i1 {{.+}}, label %{{.+}}, label %[[S_ARR_BODY]] 515 // CHECK: call {{.*}} [[ST_TY_DEFAULT_CONSTR]]([[ST_TY]]* {{[^,]*}} [[ST_TY_TEMP:%.+]]) 516 // CHECK: call {{.*}} [[S_INT_TY_COPY_CONSTR]]([[S_INT_TY]]* {{[^,]*}} [[VAR_PRIV]], [[S_INT_TY]]* {{.*}} [[VAR_REF]], [[ST_TY]]* [[ST_TY_TEMP]]) 517 // CHECK: call {{.*}} [[ST_TY_DESTR]]([[ST_TY]]* {{[^,]*}} [[ST_TY_TEMP]]) 518 // CHECK-NOT: call {{.*}}void @__kmpc_barrier( 519 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* {{[^,]*}} [[VAR_PRIV]]) 520 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* 521 // CHECK: ret void 522 523 #endif 524 #else 525 526 enum omp_allocator_handle_t { 527 omp_null_allocator = 0, 528 omp_default_mem_alloc = 1, 529 omp_large_cap_mem_alloc = 2, 530 omp_const_mem_alloc = 3, 531 omp_high_bw_mem_alloc = 4, 532 omp_low_lat_mem_alloc = 5, 533 omp_cgroup_mem_alloc = 6, 534 omp_pteam_mem_alloc = 7, 535 omp_thread_mem_alloc = 8, 536 KMP_ALLOCATOR_MAX_HANDLE = __UINTPTR_MAX__ 537 }; 538 539 struct St { 540 int a, b; 541 St() : a(0), b(0) {} 542 St(const St &) { } 543 ~St() {} 544 void St_func(St s[2], int n, long double vla1[n]) { 545 double vla2[n][n] __attribute__((aligned(128))); 546 a = b; 547 #pragma omp parallel allocate(omp_thread_mem_alloc:vla2) firstprivate(s, vla1, vla2) 548 vla1[b] = vla2[1][n - 1] = a = b; 549 } 550 }; 551 552 // ARRAY-LABEL: array_func 553 void array_func(float a[3], St s[2], int n, long double vla1[n]) { 554 double vla2[n][n] __attribute__((aligned(128))); 555 // ARRAY: @__kmpc_fork_call( 556 // ARRAY-DAG: [[PRIV_S:%.+]] = alloca %struct.St*, 557 // ARRAY-DAG: [[PRIV_VLA1:%.+]] = alloca x86_fp80*, 558 // ARRAY-DAG: [[PRIV_A:%.+]] = alloca float*, 559 // ARRAY-DAG: [[PRIV_VLA2:%.+]] = alloca double*, 560 // ARRAY-DAG: store %struct.St* %{{.+}}, %struct.St** [[PRIV_S]], 561 // ARRAY-DAG: store x86_fp80* %{{.+}}, x86_fp80** [[PRIV_VLA1]], 562 // ARRAY-DAG: store float* %{{.+}}, float** [[PRIV_A]], 563 // ARRAY-DAG: store double* %{{.+}}, double** [[PRIV_VLA2]], 564 // ARRAY: call i8* @llvm.stacksave() 565 // ARRAY: [[SIZE:%.+]] = mul nuw i64 %{{.+}}, 8 566 // ARRAY: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 128 %{{.+}}, i8* align 128 %{{.+}}, i64 [[SIZE]], i1 false) 567 #pragma omp parallel firstprivate(a, s, vla1, vla2) 568 s[0].St_func(s, n, vla1); 569 ; 570 } 571 572 // ARRAY-LABEL: St_func 573 // ARRAY: @__kmpc_fork_call( 574 // ARRAY-DAG: [[PRIV_VLA1:%.+]] = alloca x86_fp80*, 575 // ARRAY-DAG: [[PRIV_S:%.+]] = alloca %struct.St*, 576 // ARRAY-DAG: [[PRIV_VLA2:%.+]] = alloca double*, 577 // ARRAY-DAG: store %struct.St* %{{.+}}, %struct.St** [[PRIV_S]], 578 // ARRAY-DAG: store x86_fp80* %{{.+}}, x86_fp80** [[PRIV_VLA1]], 579 // ARRAY-DAG: store double* %{{.+}}, double** [[PRIV_VLA2]], 580 // ARRAY: [[SIZE:%.+]] = mul nuw i64 %{{.+}}, 8 581 // ARRAY: [[SZ1:%.+]] = add nuw i64 [[SIZE]], 127 582 // ARRAY: [[SZ2:%.+]] = udiv i64 [[SZ1]], 128 583 // ARRAY: [[SIZE:%.+]] = mul nuw i64 [[SZ2]], 128 584 // ARRAY: [[VLA2_VOID_PTR:%.+]] = call i8* @__kmpc_alloc(i32 [[GTID:%.+]], i64 [[SIZE]], i8* inttoptr (i64 8 to i8*)) 585 // ARRAY: [[VLA2_PTR:%.+]] = bitcast i8* [[VLA2_VOID_PTR]] to double* 586 // ARRAY: [[SIZE:%.+]] = mul nuw i64 %{{.+}}, 8 587 // ARRAY: [[BC:%.+]] = bitcast double* [[VLA2_PTR]] to i8* 588 // ARRAY: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 128 [[BC]], i8* align 128 %{{.+}}, i64 [[SIZE]], i1 false) 589 // ARRAY: [[VLA2_VOID_PTR:%.+]] = bitcast double* [[VLA2_PTR]] to i8* 590 // ARRAY: call void @__kmpc_free(i32 [[GTID]], i8* [[VLA2_VOID_PTR]], i8* inttoptr (i64 8 to i8*)) 591 // ARRAY-NEXT: ret void 592 #endif 593 594 595