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