1 // RUN: %clang_cc1 -verify -fopenmp=libiomp5 -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck %s 2 // RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s 3 // RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s 4 // RUN: %clang_cc1 -verify -fopenmp=libiomp5 -x c++ -std=c++11 -DLAMBDA -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -check-prefix=LAMBDA %s 5 // RUN: %clang_cc1 -verify -fopenmp=libiomp5 -x c++ -fblocks -DBLOCKS -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -check-prefix=BLOCKS %s 6 // expected-no-diagnostics 7 #ifndef HEADER 8 #define HEADER 9 10 volatile double g; 11 12 template <class T> 13 struct S { 14 T f; 15 S(T a) : f(a + g) {} 16 S() : f(g) {} 17 operator T() { return T(); } 18 S &operator&(const S &) { return *this; } 19 ~S() {} 20 }; 21 22 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { float } 23 // CHECK-DAG: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} } 24 // CHECK-DAG: [[CAP_MAIN_TY:%.+]] = type { float*, [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]*, float*, [2 x i{{[0-9]+}}]*, [2 x [[S_FLOAT_TY]]]* } 25 // CHECK-DAG: [[CAP_TMAIN_TY:%.+]] = type { i{{[0-9]+}}*, [[S_INT_TY]]*, [[S_INT_TY]]*, i{{[0-9]+}}*, [2 x i{{[0-9]+}}]*, [2 x [[S_INT_TY]]]* } 26 // CHECK-DAG: [[ATOMIC_REDUCE_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 27 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8* 28 // CHECK-DAG: [[SINGLE_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 322, i32 0, i32 0, i8* 29 // CHECK-DAG: [[REDUCTION_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 30 // CHECK-DAG: [[REDUCTION_LOCK:@.+]] = common global [8 x i32] zeroinitializer 31 32 template <typename T> 33 T tmain() { 34 T t; 35 S<T> test; 36 T t_var = T(), t_var1; 37 T vec[] = {1, 2}; 38 S<T> s_arr[] = {1, 2}; 39 S<T> var(3), var1; 40 #pragma omp parallel 41 #pragma omp sections reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) nowait 42 { 43 vec[0] = t_var; 44 #pragma omp section 45 s_arr[0] = var; 46 } 47 return T(); 48 } 49 50 int main() { 51 #ifdef LAMBDA 52 // LAMBDA: [[G:@.+]] = global double 53 // LAMBDA-LABEL: @main 54 // LAMBDA: call void [[OUTER_LAMBDA:@.+]]( 55 [&]() { 56 // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]]( 57 // LAMBDA: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 1, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}, i8* %{{.+}}) 58 #pragma omp parallel 59 #pragma omp sections reduction(+:g) 60 { 61 // LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* %{{.+}}, i32* %{{.+}}, %{{.+}}* %{{.+}}) 62 // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca double, 63 64 // Reduction list for runtime. 65 // LAMBDA: [[RED_LIST:%.+]] = alloca [1 x i8*], 66 67 // LAMBDA: store double 0.0{{.+}}, double* [[G_PRIVATE_ADDR]] 68 // LAMBDA: call void @__kmpc_for_static_init_4( 69 g = 1; 70 // LAMBDA: store volatile double 1.0{{.+}}, double* [[G_PRIVATE_ADDR]], 71 // LAMBDA: [[G_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 72 // LAMBDA: store double* [[G_PRIVATE_ADDR]], double** [[G_PRIVATE_ADDR_REF]] 73 // LAMBDA: call void [[INNER_LAMBDA:@.+]](%{{.+}}* [[ARG]]) 74 // LAMBDA: call void @__kmpc_for_static_fini( 75 76 // LAMBDA: [[G_PRIV_REF:%.+]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[RED_LIST]], i32 0, i32 0 77 // LAMBDA: [[BITCAST:%.+]] = bitcast double* [[G_PRIVATE_ADDR]] to i8* 78 // LAMBDA: store i8* [[BITCAST]], i8** [[G_PRIV_REF]], 79 // LAMBDA: call i32 @__kmpc_reduce( 80 // LAMBDA: switch i32 %{{.+}}, label %[[REDUCTION_DONE:.+]] [ 81 // LAMBDA: i32 1, label %[[CASE1:.+]] 82 // LAMBDA: i32 2, label %[[CASE2:.+]] 83 // LAMBDA: [[CASE1]] 84 // LAMBDA: [[G_VAL:%.+]] = load double, double* [[G]] 85 // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 86 // LAMBDA: [[ADD:%.+]] = fadd double [[G_VAL]], [[G_PRIV_VAL]] 87 // LAMBDA: store double [[ADD]], double* [[G]] 88 // LAMBDA: call void @__kmpc_end_reduce( 89 // LAMBDA: br label %[[REDUCTION_DONE]] 90 // LAMBDA: [[CASE2]] 91 // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 92 // LAMBDA: fadd double 93 // LAMBDA: cmpxchg i64* 94 // LAMBDA: br label %[[REDUCTION_DONE]] 95 // LAMBDA: [[REDUCTION_DONE]] 96 // LAMBDA: ret void 97 #pragma omp section 98 [&]() { 99 // LAMBDA: define {{.+}} void [[INNER_LAMBDA]](%{{.+}}* [[ARG_PTR:%.+]]) 100 // LAMBDA: store %{{.+}}* [[ARG_PTR]], %{{.+}}** [[ARG_PTR_REF:%.+]], 101 g = 2; 102 // LAMBDA: [[ARG_PTR:%.+]] = load %{{.+}}*, %{{.+}}** [[ARG_PTR_REF]] 103 // LAMBDA: [[G_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 104 // LAMBDA: [[G_REF:%.+]] = load double*, double** [[G_PTR_REF]] 105 // LAMBDA: store volatile double 2.0{{.+}}, double* [[G_REF]] 106 }(); 107 } 108 }(); 109 return 0; 110 #elif defined(BLOCKS) 111 // BLOCKS: [[G:@.+]] = global double 112 // BLOCKS-LABEL: @main 113 // BLOCKS: call void {{%.+}}(i8 114 ^{ 115 // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8* 116 // BLOCKS: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 1, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}, i8* %{{.+}}) 117 #pragma omp parallel 118 #pragma omp sections reduction(-:g) 119 { 120 // BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* %{{.+}}, i32* %{{.+}}, %{{.+}}* %{{.+}}) 121 // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca double, 122 123 // Reduction list for runtime. 124 // BLOCKS: [[RED_LIST:%.+]] = alloca [1 x i8*], 125 126 // BLOCKS: store double 0.0{{.+}}, double* [[G_PRIVATE_ADDR]] 127 g = 1; 128 // BLOCKS: call void @__kmpc_for_static_init_4( 129 // BLOCKS: store volatile double 1.0{{.+}}, double* [[G_PRIVATE_ADDR]], 130 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 131 // BLOCKS: double* [[G_PRIVATE_ADDR]] 132 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 133 // BLOCKS: call void {{%.+}}(i8 134 // BLOCKS: call void @__kmpc_for_static_fini( 135 136 // BLOCKS: [[G_PRIV_REF:%.+]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[RED_LIST]], i32 0, i32 0 137 // BLOCKS: [[BITCAST:%.+]] = bitcast double* [[G_PRIVATE_ADDR]] to i8* 138 // BLOCKS: store i8* [[BITCAST]], i8** [[G_PRIV_REF]], 139 // BLOCKS: call i32 @__kmpc_reduce( 140 // BLOCKS: switch i32 %{{.+}}, label %[[REDUCTION_DONE:.+]] [ 141 // BLOCKS: i32 1, label %[[CASE1:.+]] 142 // BLOCKS: i32 2, label %[[CASE2:.+]] 143 // BLOCKS: [[CASE1]] 144 // BLOCKS: [[G_VAL:%.+]] = load double, double* [[G]] 145 // BLOCKS: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 146 // BLOCKS: [[ADD:%.+]] = fadd double [[G_VAL]], [[G_PRIV_VAL]] 147 // BLOCKS: store double [[ADD]], double* [[G]] 148 // BLOCKS: call void @__kmpc_end_reduce( 149 // BLOCKS: br label %[[REDUCTION_DONE]] 150 // BLOCKS: [[CASE2]] 151 // BLOCKS: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 152 // BLOCKS: fadd double 153 // BLOCKS: cmpxchg i64* 154 // BLOCKS: br label %[[REDUCTION_DONE]] 155 // BLOCKS: [[REDUCTION_DONE]] 156 // BLOCKS: ret void 157 #pragma omp section 158 ^{ 159 // BLOCKS: define {{.+}} void {{@.+}}(i8* 160 g = 2; 161 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 162 // BLOCKS: store volatile double 2.0{{.+}}, double* 163 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 164 // BLOCKS: ret 165 }(); 166 } 167 }(); 168 return 0; 169 #else 170 S<float> test; 171 float t_var = 0, t_var1; 172 int vec[] = {1, 2}; 173 S<float> s_arr[] = {1, 2}; 174 S<float> var(3), var1; 175 #pragma omp parallel 176 #pragma omp sections reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) 177 { 178 { 179 vec[0] = t_var; 180 s_arr[0] = var; 181 vec[1] = t_var1; 182 s_arr[1] = var1; 183 } 184 } 185 return tmain<int>(); 186 #endif 187 } 188 189 // CHECK: define {{.*}}i{{[0-9]+}} @main() 190 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]], 191 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]]) 192 // CHECK: %{{.+}} = bitcast [[CAP_MAIN_TY]]* 193 // 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]+}}*, [[CAP_MAIN_TY]]*)* [[MAIN_MICROTASK:@.+]] to void 194 // CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT:@.+]]() 195 // CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]* 196 // CHECK: ret 197 // 198 // CHECK: define internal void [[MAIN_MICROTASK]](i{{[0-9]+}}* [[GTID_ADDR:%.+]], i{{[0-9]+}}* %{{.+}}, [[CAP_MAIN_TY]]* %{{.+}}) 199 // CHECK-NOT: alloca float, 200 // CHECK-NOT: alloca [[S_FLOAT_TY]], 201 // CHECK-NOT: alloca [[S_FLOAT_TY]], 202 // CHECK-NOT: alloca float, 203 204 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 205 206 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 207 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 208 // CHECK: call i32 @__kmpc_single( 209 210 // CHECK-DAG: getelementptr inbounds [[CAP_MAIN_TY]], [[CAP_MAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} 0 211 // CHECK-DAG: getelementptr inbounds [[CAP_MAIN_TY]], [[CAP_MAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} 1 212 // CHECK-DAG: getelementptr inbounds [[CAP_MAIN_TY]], [[CAP_MAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} 2 213 // CHECK-DAG: getelementptr inbounds [[CAP_MAIN_TY]], [[CAP_MAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} 3 214 215 // CHECK-NOT: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 216 // CHECK-NOT: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* 217 218 // CHECK: call void @__kmpc_end_single( 219 220 // CHECK: call i32 @__kmpc_cancel_barrier(%{{.+}}* [[SINGLE_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 221 // CHECK: call i32 @__kmpc_cancel_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 222 223 // CHECK: ret void 224 225 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT]]() 226 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]], 227 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]]) 228 // CHECK: %{{.+}} = bitcast [[CAP_TMAIN_TY]]* 229 // 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]+}}*, [[CAP_TMAIN_TY]]*)* [[TMAIN_MICROTASK:@.+]] to void 230 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]* 231 // CHECK: ret 232 // 233 // CHECK: define internal void [[TMAIN_MICROTASK]](i{{[0-9]+}}* [[GTID_ADDR:%.+]], i{{[0-9]+}}* %{{.+}}, [[CAP_TMAIN_TY]]* %{{.+}}) 234 // CHECK: alloca i{{[0-9]+}}, 235 // CHECK: alloca i{{[0-9]+}}, 236 // CHECK: alloca i{{[0-9]+}}, 237 // CHECK: alloca i{{[0-9]+}}, 238 // CHECK: alloca i{{[0-9]+}}, 239 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 240 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], 241 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_INT_TY]], 242 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca i{{[0-9]+}}, 243 244 // Reduction list for runtime. 245 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 246 247 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 248 249 // CHECK: [[T_VAR_PTR_REF:%.+]] = getelementptr inbounds [[CAP_TMAIN_TY]], [[CAP_TMAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} {{[0-9]+}} 250 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[T_VAR_PTR_REF]], 251 // For + reduction operation initial value of private variable is 0. 252 // CHECK: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[T_VAR_PRIV]], 253 254 // CHECK: [[VAR_PTR_REF:%.+]] = getelementptr inbounds [[CAP_TMAIN_TY]], [[CAP_TMAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} {{[0-9]+}} 255 // CHECK: [[VAR_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** [[VAR_PTR_REF:%.+]], 256 // For & reduction operation initial value of private variable is ones in all bits. 257 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR_PRIV]]) 258 259 // CHECK: [[VAR1_PTR_REF:%.+]] = getelementptr inbounds [[CAP_TMAIN_TY]], [[CAP_TMAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} {{[0-9]+}} 260 // CHECK: [[VAR1_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** [[VAR_PTR_REF:%.+]], 261 // For && reduction operation initial value of private variable is 1.0. 262 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR1_PRIV]]) 263 264 // CHECK: [[T_VAR1_PTR_REF:%.+]] = getelementptr inbounds [[CAP_TMAIN_TY]], [[CAP_TMAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} {{[0-9]+}} 265 // CHECK: [[T_VAR1_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[T_VAR1_PTR_REF]], 266 // For min reduction operation initial value of private variable is largest repesentable value. 267 // CHECK: store i{{[0-9]+}} 2147483647, i{{[0-9]+}}* [[T_VAR1_PRIV]], 268 269 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 270 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 271 // CHECK: call void @__kmpc_for_static_init_4( 272 // Skip checks for internal operations. 273 // CHECK: call void @__kmpc_for_static_fini( 274 275 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 276 277 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i32 0, i32 0 278 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR_PRIV]] to i8* 279 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 280 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i32 0, i32 1 281 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR_PRIV]] to i8* 282 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 283 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i32 0, i32 2 284 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_PRIV]] to i8* 285 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 286 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i32 0, i32 3 287 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR1_PRIV]] to i8* 288 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 289 290 // res = __kmpc_reduce_nowait(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 291 292 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 293 // CHECK: [[RES:%.+]] = call i32 @__kmpc_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], i32 4, i64 32, i8* [[BITCAST]], void (i8*, i8*)* [[REDUCTION_FUNC:@.+]], [8 x i32]* [[REDUCTION_LOCK]]) 294 295 // switch(res) 296 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 297 // CHECK: i32 1, label %[[CASE1:.+]] 298 // CHECK: i32 2, label %[[CASE2:.+]] 299 // CHECK: ] 300 301 // case 1: 302 // t_var += t_var_reduction; 303 // CHECK: [[T_VAR_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_REF]], 304 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]], 305 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_VAL]], [[T_VAR_PRIV_VAL]] 306 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_REF]], 307 308 // var = var.operator &(var_reduction); 309 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 310 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* 311 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 312 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 313 314 // var1 = var1.operator &&(var1_reduction); 315 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) 316 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 317 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[FALSE:.+]] 318 // CHECK: [[TRUE]] 319 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) 320 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 321 // CHECK: br i1 [[VAR1_REDUCTION_BOOL]], label %[[TRUE2:.+]], label %[[FALSE2:.+]] 322 // CHECK: [[TRUE2]] 323 // CHECK: br label %[[END2:.+]] 324 // CHECK: [[FALSE2]] 325 // CHECK: br label %[[END2]] 326 // CHECK: [[END2]] 327 // CHECK: [[COND_LVALUE:%.+]] = phi [[S_INT_TY]]* [ [[VAR1_REF]], %[[TRUE2]] ], [ [[VAR1_PRIV]], %[[FALSE2]] ] 328 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* 329 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 330 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 331 332 // t_var1 = min(t_var1, t_var1_reduction); 333 // CHECK: [[T_VAR1_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_REF]], 334 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]], 335 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_VAL]], [[T_VAR1_PRIV_VAL]] 336 // CHECK: [[UP:%.+]] = zext i1 [[CMP]] to i{{[0-9]+}} 337 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_REF]], 338 339 // __kmpc_end_reduce_nowait(<loc>, <gtid>, &<lock>); 340 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 341 342 // break; 343 // CHECK: br label %[[RED_DONE]] 344 345 // case 2: 346 // t_var += t_var_reduction; 347 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]] 348 // CHECK: atomicrmw add i32* [[T_VAR_REF]], i32 [[T_VAR_PRIV_VAL]] monotonic 349 350 // var = var.operator &(var_reduction); 351 // CHECK: call void @__kmpc_critical( 352 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 353 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* 354 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 355 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 356 // CHECK: call void @__kmpc_end_critical( 357 358 // var1 = var1.operator &&(var1_reduction); 359 // CHECK: call void @__kmpc_critical( 360 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) 361 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 362 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[FALSE:.+]] 363 // CHECK: [[TRUE]] 364 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) 365 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 366 // CHECK: br i1 [[VAR1_REDUCTION_BOOL]], label %[[TRUE2:.+]], label %[[FALSE2:.+]] 367 // CHECK: [[TRUE2]] 368 // CHECK: br label %[[END2:.+]] 369 // CHECK: [[FALSE2]] 370 // CHECK: br label %[[END2]] 371 // CHECK: [[END2]] 372 // CHECK: [[COND_LVALUE:%.+]] = phi [[S_INT_TY]]* [ [[VAR1_REF]], %[[TRUE2]] ], [ [[VAR1_PRIV]], %[[FALSE2]] ] 373 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* 374 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 375 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 376 // CHECK: call void @__kmpc_end_critical( 377 378 // t_var1 = min(t_var1, t_var1_reduction); 379 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]] 380 // CHECK: atomicrmw min i32* [[T_VAR1_REF]], i32 [[T_VAR1_PRIV_VAL]] monotonic 381 382 // break; 383 // CHECK: br label %[[RED_DONE]] 384 // CHECK: [[RED_DONE]] 385 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) 386 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* 387 // CHECK: call i32 @__kmpc_cancel_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 388 // CHECK: ret void 389 390 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 391 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 392 // ... 393 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 394 // *(Type<n>-1*)rhs[<n>-1]); 395 // } 396 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 397 // t_var_lhs = (i{{[0-9]+}}*)lhs[0]; 398 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i32 0, i32 0 399 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 400 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to i{{[0-9]+}}* 401 // t_var_rhs = (i{{[0-9]+}}*)rhs[0]; 402 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i32 0, i32 0 403 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 404 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to i{{[0-9]+}}* 405 406 // var_lhs = (S<i{{[0-9]+}}>*)lhs[1]; 407 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i32 0, i32 1 408 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 409 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_INT_TY]]* 410 // var_rhs = (S<i{{[0-9]+}}>*)rhs[1]; 411 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i32 0, i32 1 412 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 413 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_INT_TY]]* 414 415 // var1_lhs = (S<i{{[0-9]+}}>*)lhs[2]; 416 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i32 0, i32 2 417 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 418 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_INT_TY]]* 419 // var1_rhs = (S<i{{[0-9]+}}>*)rhs[2]; 420 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i32 0, i32 2 421 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 422 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_INT_TY]]* 423 424 // t_var1_lhs = (i{{[0-9]+}}*)lhs[3]; 425 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i32 0, i32 3 426 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 427 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to i{{[0-9]+}}* 428 // t_var1_rhs = (i{{[0-9]+}}*)rhs[3]; 429 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i32 0, i32 3 430 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 431 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to i{{[0-9]+}}* 432 433 // t_var_lhs += t_var_rhs; 434 // CHECK: [[T_VAR_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_LHS]], 435 // CHECK: [[T_VAR_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_RHS]], 436 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_LHS_VAL]], [[T_VAR_RHS_VAL]] 437 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_LHS]], 438 439 // var_lhs = var_lhs.operator &(var_rhs); 440 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_LHS]], [[S_INT_TY]]* dereferenceable(4) [[VAR_RHS]]) 441 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_LHS]] to i8* 442 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 443 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 444 445 // var1_lhs = var1_lhs.operator &&(var1_rhs); 446 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_LHS]]) 447 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 448 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[FALSE:.+]] 449 // CHECK: [[TRUE]] 450 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_RHS]]) 451 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 452 // CHECK: br i1 [[VAR1_REDUCTION_BOOL]], label %[[TRUE2:.+]], label %[[FALSE2:.+]] 453 // CHECK: [[TRUE2]] 454 // CHECK: br label %[[END2:.+]] 455 // CHECK: [[FALSE2]] 456 // CHECK: br label %[[END2]] 457 // CHECK: [[END2]] 458 // CHECK: [[COND_LVALUE:%.+]] = phi [[S_INT_TY]]* [ [[VAR1_LHS]], %[[TRUE2]] ], [ [[VAR1_RHS]], %[[FALSE2]] ] 459 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_LHS]] to i8* 460 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 461 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 462 463 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 464 // CHECK: [[T_VAR1_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_LHS]], 465 // CHECK: [[T_VAR1_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_RHS]], 466 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_LHS_VAL]], [[T_VAR1_RHS_VAL]] 467 // CHECK: [[UP:%.+]] = zext i1 [[CMP]] to i{{[0-9]+}} 468 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_LHS]], 469 // CHECK: ret void 470 471 #endif 472 473