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 // REQUIRES: x86-registered-target 8 #ifndef HEADER 9 #define HEADER 10 11 volatile double g; 12 13 template <class T> 14 struct S { 15 T f; 16 S(T a) : f(a + g) {} 17 S() : f(g) {} 18 operator T() { return T(); } 19 S &operator&(const S &) { return *this; } 20 ~S() {} 21 }; 22 23 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { float } 24 // CHECK-DAG: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} } 25 // CHECK-DAG: [[ATOMIC_REDUCE_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 26 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8* 27 // CHECK-DAG: [[SINGLE_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 322, i32 0, i32 0, i8* 28 // CHECK-DAG: [[REDUCTION_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 29 // CHECK-DAG: [[REDUCTION_LOCK:@.+]] = common global [8 x i32] zeroinitializer 30 31 template <typename T> 32 T tmain() { 33 T t; 34 S<T> test; 35 T t_var = T(), t_var1; 36 T vec[] = {1, 2}; 37 S<T> s_arr[] = {1, 2}; 38 S<T> var(3), var1; 39 #pragma omp parallel 40 #pragma omp sections reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) nowait 41 { 42 vec[0] = t_var; 43 #pragma omp section 44 s_arr[0] = var; 45 } 46 return T(); 47 } 48 49 int main() { 50 #ifdef LAMBDA 51 // LAMBDA: [[G:@.+]] = global double 52 // LAMBDA-LABEL: @main 53 // LAMBDA: call void [[OUTER_LAMBDA:@.+]]( 54 [&]() { 55 // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]]( 56 // LAMBDA: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}) 57 #pragma omp parallel 58 #pragma omp sections reduction(+:g) 59 { 60 // LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}) 61 // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca double, 62 63 // Reduction list for runtime. 64 // LAMBDA: [[RED_LIST:%.+]] = alloca [1 x i8*], 65 66 // LAMBDA: store double 0.0{{.+}}, double* [[G_PRIVATE_ADDR]] 67 // LAMBDA: call void @__kmpc_for_static_init_4( 68 g = 1; 69 // LAMBDA: store double 1.0{{.+}}, double* [[G_PRIVATE_ADDR]], 70 // LAMBDA: [[G_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 71 // LAMBDA: store double* [[G_PRIVATE_ADDR]], double** [[G_PRIVATE_ADDR_REF]] 72 // LAMBDA: call void [[INNER_LAMBDA:@.+]](%{{.+}}* [[ARG]]) 73 // LAMBDA: call void @__kmpc_for_static_fini( 74 75 // LAMBDA: [[G_PRIV_REF:%.+]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[RED_LIST]], i64 0, i64 0 76 // LAMBDA: [[BITCAST:%.+]] = bitcast double* [[G_PRIVATE_ADDR]] to i8* 77 // LAMBDA: store i8* [[BITCAST]], i8** [[G_PRIV_REF]], 78 // LAMBDA: call i32 @__kmpc_reduce( 79 // LAMBDA: switch i32 %{{.+}}, label %[[REDUCTION_DONE:.+]] [ 80 // LAMBDA: i32 1, label %[[CASE1:.+]] 81 // LAMBDA: i32 2, label %[[CASE2:.+]] 82 // LAMBDA: [[CASE1]] 83 // LAMBDA: [[G_VAL:%.+]] = load double, double* [[G]] 84 // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 85 // LAMBDA: [[ADD:%.+]] = fadd double [[G_VAL]], [[G_PRIV_VAL]] 86 // LAMBDA: store double [[ADD]], double* [[G]] 87 // LAMBDA: call void @__kmpc_end_reduce( 88 // LAMBDA: br label %[[REDUCTION_DONE]] 89 // LAMBDA: [[CASE2]] 90 // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 91 // LAMBDA: fadd double 92 // LAMBDA: cmpxchg i64* 93 // LAMBDA: call void @__kmpc_end_reduce( 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 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 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}) 117 #pragma omp parallel 118 #pragma omp sections reduction(-:g) 119 { 120 // BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}) 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 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]], i64 0, i64 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: call void @__kmpc_end_reduce( 155 // BLOCKS: br label %[[REDUCTION_DONE]] 156 // BLOCKS: [[REDUCTION_DONE]] 157 // BLOCKS: ret void 158 #pragma omp section 159 ^{ 160 // BLOCKS: define {{.+}} void {{@.+}}(i8* 161 g = 2; 162 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 163 // BLOCKS: store double 2.0{{.+}}, double* 164 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 165 // BLOCKS: ret 166 }(); 167 } 168 }(); 169 return 0; 170 #else 171 S<float> test; 172 float t_var = 0, t_var1; 173 int vec[] = {1, 2}; 174 S<float> s_arr[] = {1, 2}; 175 S<float> var(3), var1; 176 #pragma omp parallel 177 #pragma omp sections reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) 178 { 179 { 180 vec[0] = t_var; 181 s_arr[0] = var; 182 vec[1] = t_var1; 183 s_arr[1] = var1; 184 } 185 } 186 return tmain<int>(); 187 #endif 188 } 189 190 // CHECK: define {{.*}}i{{[0-9]+}} @main() 191 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]], 192 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]]) 193 // 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 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]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, 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-NOT: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 211 // CHECK-NOT: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* 212 213 // CHECK: call void @__kmpc_end_single( 214 215 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[SINGLE_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 216 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 217 218 // CHECK: ret void 219 220 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT]]() 221 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]], 222 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]]) 223 // 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 224 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]* 225 // CHECK: ret 226 // 227 // CHECK: define internal void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, 228 // CHECK: alloca i{{[0-9]+}}, 229 // CHECK: alloca i{{[0-9]+}}, 230 // CHECK: alloca i{{[0-9]+}}, 231 // CHECK: alloca i{{[0-9]+}}, 232 // CHECK: alloca i{{[0-9]+}}, 233 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 234 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], 235 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_INT_TY]], 236 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca i{{[0-9]+}}, 237 238 // Reduction list for runtime. 239 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 240 241 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 242 243 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 244 // CHECK: [[VAR_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 245 // CHECK: [[VAR1_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 246 // CHECK: [[T_VAR1_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 247 248 // For + reduction operation initial value of private variable is 0. 249 // CHECK: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[T_VAR_PRIV]], 250 251 // For & reduction operation initial value of private variable is ones in all bits. 252 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR_PRIV]]) 253 254 // For && reduction operation initial value of private variable is 1.0. 255 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR1_PRIV]]) 256 257 // For min reduction operation initial value of private variable is largest repesentable value. 258 // CHECK: store i{{[0-9]+}} 2147483647, i{{[0-9]+}}* [[T_VAR1_PRIV]], 259 260 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 261 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 262 // CHECK: call void @__kmpc_for_static_init_4( 263 // Skip checks for internal operations. 264 // CHECK: call void @__kmpc_for_static_fini( 265 266 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 267 268 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 269 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR_PRIV]] to i8* 270 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 271 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 272 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR_PRIV]] to i8* 273 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 274 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 275 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_PRIV]] to i8* 276 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 277 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 278 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR1_PRIV]] to i8* 279 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 280 281 // res = __kmpc_reduce_nowait(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 282 283 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 284 // 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]]) 285 286 // switch(res) 287 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 288 // CHECK: i32 1, label %[[CASE1:.+]] 289 // CHECK: i32 2, label %[[CASE2:.+]] 290 // CHECK: ] 291 292 // case 1: 293 // t_var += t_var_reduction; 294 // CHECK: [[T_VAR_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_REF]], 295 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]], 296 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_VAL]], [[T_VAR_PRIV_VAL]] 297 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_REF]], 298 299 // var = var.operator &(var_reduction); 300 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 301 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* 302 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 303 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 304 305 // var1 = var1.operator &&(var1_reduction); 306 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) 307 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 308 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 309 // CHECK: [[TRUE]] 310 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) 311 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 312 // CHECK: br label %[[END2]] 313 // CHECK: [[END2]] 314 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 315 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 316 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 317 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* 318 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 319 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 320 321 // t_var1 = min(t_var1, t_var1_reduction); 322 // CHECK: [[T_VAR1_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_REF]], 323 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]], 324 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_VAL]], [[T_VAR1_PRIV_VAL]] 325 // CHECK: br i1 [[CMP]] 326 // CHECK: [[UP:%.+]] = phi i32 327 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_REF]], 328 329 // __kmpc_end_reduce_nowait(<loc>, <gtid>, &<lock>); 330 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 331 332 // break; 333 // CHECK: br label %[[RED_DONE]] 334 335 // case 2: 336 // t_var += t_var_reduction; 337 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]] 338 // CHECK: atomicrmw add i32* [[T_VAR_REF]], i32 [[T_VAR_PRIV_VAL]] monotonic 339 340 // var = var.operator &(var_reduction); 341 // CHECK: call void @__kmpc_critical( 342 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 343 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* 344 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 345 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 346 // CHECK: call void @__kmpc_end_critical( 347 348 // var1 = var1.operator &&(var1_reduction); 349 // CHECK: call void @__kmpc_critical( 350 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) 351 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 352 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 353 // CHECK: [[TRUE]] 354 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) 355 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 356 // CHECK: br label %[[END2]] 357 // CHECK: [[END2]] 358 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 359 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 360 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 361 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* 362 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 363 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 364 // CHECK: call void @__kmpc_end_critical( 365 366 // t_var1 = min(t_var1, t_var1_reduction); 367 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]] 368 // CHECK: atomicrmw min i32* [[T_VAR1_REF]], i32 [[T_VAR1_PRIV_VAL]] monotonic 369 370 // break; 371 // CHECK: br label %[[RED_DONE]] 372 // CHECK: [[RED_DONE]] 373 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) 374 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* 375 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 376 // CHECK: ret void 377 378 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 379 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 380 // ... 381 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 382 // *(Type<n>-1*)rhs[<n>-1]); 383 // } 384 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 385 // t_var_lhs = (i{{[0-9]+}}*)lhs[0]; 386 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 387 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 388 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to i{{[0-9]+}}* 389 // t_var_rhs = (i{{[0-9]+}}*)rhs[0]; 390 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 391 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 392 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to i{{[0-9]+}}* 393 394 // var_lhs = (S<i{{[0-9]+}}>*)lhs[1]; 395 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 396 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 397 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_INT_TY]]* 398 // var_rhs = (S<i{{[0-9]+}}>*)rhs[1]; 399 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 400 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 401 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_INT_TY]]* 402 403 // var1_lhs = (S<i{{[0-9]+}}>*)lhs[2]; 404 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 405 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 406 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_INT_TY]]* 407 // var1_rhs = (S<i{{[0-9]+}}>*)rhs[2]; 408 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 409 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 410 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_INT_TY]]* 411 412 // t_var1_lhs = (i{{[0-9]+}}*)lhs[3]; 413 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 414 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 415 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to i{{[0-9]+}}* 416 // t_var1_rhs = (i{{[0-9]+}}*)rhs[3]; 417 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 418 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 419 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to i{{[0-9]+}}* 420 421 // t_var_lhs += t_var_rhs; 422 // CHECK: [[T_VAR_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_LHS]], 423 // CHECK: [[T_VAR_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_RHS]], 424 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_LHS_VAL]], [[T_VAR_RHS_VAL]] 425 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_LHS]], 426 427 // var_lhs = var_lhs.operator &(var_rhs); 428 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_LHS]], [[S_INT_TY]]* dereferenceable(4) [[VAR_RHS]]) 429 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_LHS]] to i8* 430 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 431 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 432 433 // var1_lhs = var1_lhs.operator &&(var1_rhs); 434 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_LHS]]) 435 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 436 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 437 // CHECK: [[TRUE]] 438 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_RHS]]) 439 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 440 // CHECK: br label %[[END2]] 441 // CHECK: [[END2]] 442 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 443 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 444 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 445 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_LHS]] to i8* 446 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 447 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 448 449 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 450 // CHECK: [[T_VAR1_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_LHS]], 451 // CHECK: [[T_VAR1_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_RHS]], 452 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_LHS_VAL]], [[T_VAR1_RHS_VAL]] 453 // CHECK: br i1 [[CMP]] 454 // CHECK: [[UP:%.+]] = phi i32 455 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_LHS]], 456 // CHECK: ret void 457 458 #endif 459 460