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