1 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap %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 -allow-deprecated-dag-overlap %s 4 // RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DLAMBDA -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap -check-prefix=LAMBDA %s 5 // RUN: %clang_cc1 -verify -fopenmp -x c++ -fblocks -DBLOCKS -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap -check-prefix=BLOCKS %s 6 7 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap --check-prefix SIMD-ONLY0 %s 8 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s 9 // RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap --check-prefix SIMD-ONLY0 %s 10 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -std=c++11 -DLAMBDA -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap --check-prefix SIMD-ONLY0 %s 11 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -fblocks -DBLOCKS -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap --check-prefix SIMD-ONLY0 %s 12 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}} 13 // expected-no-diagnostics 14 #ifndef HEADER 15 #define HEADER 16 17 volatile double g; 18 19 template <class T> 20 struct S { 21 T f; 22 S(T a) : f(a + g) {} 23 S() : f(g) {} 24 operator T() { return T(); } 25 S &operator&(const S &) { return *this; } 26 ~S() {} 27 }; 28 29 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { float } 30 // CHECK-DAG: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} } 31 // CHECK-DAG: [[ATOMIC_REDUCE_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 32 // CHECK-DAG: [[REDUCTION_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 33 // CHECK-DAG: [[REDUCTION_LOCK:@.+]] = common global [8 x i32] zeroinitializer 34 35 template <typename T> 36 T tmain() { 37 T t; 38 S<T> test; 39 T t_var = T(), t_var1; 40 T vec[] = {1, 2}; 41 S<T> s_arr[] = {1, 2}; 42 S<T> var(3), var1; 43 #pragma omp parallel 44 #pragma omp sections reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) nowait 45 { 46 vec[0] = t_var; 47 #pragma omp section 48 s_arr[0] = var; 49 } 50 return T(); 51 } 52 53 int main() { 54 #ifdef LAMBDA 55 // LAMBDA: [[G:@.+]] = global double 56 // LAMBDA-LABEL: @main 57 // LAMBDA: call void [[OUTER_LAMBDA:@.+]]( 58 [&]() { 59 // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]]( 60 // LAMBDA: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}) 61 #pragma omp parallel 62 #pragma omp sections reduction(+:g) 63 { 64 // LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}) 65 // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca double, 66 67 // Reduction list for runtime. 68 // LAMBDA: [[RED_LIST:%.+]] = alloca [1 x i8*], 69 70 // LAMBDA: store double 0.0{{.+}}, double* [[G_PRIVATE_ADDR]] 71 // LAMBDA: call void @__kmpc_for_static_init_4( 72 g = 1; 73 // LAMBDA: store double 1.0{{.+}}, double* [[G_PRIVATE_ADDR]], 74 // LAMBDA: [[G_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 75 // LAMBDA: store double* [[G_PRIVATE_ADDR]], double** [[G_PRIVATE_ADDR_REF]] 76 // LAMBDA: call void [[INNER_LAMBDA:@.+]](%{{.+}}* [[ARG]]) 77 // LAMBDA: call void @__kmpc_for_static_fini( 78 79 // LAMBDA: [[G_PRIV_REF:%.+]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[RED_LIST]], i64 0, i64 0 80 // LAMBDA: [[BITCAST:%.+]] = bitcast double* [[G_PRIVATE_ADDR]] to i8* 81 // LAMBDA: store i8* [[BITCAST]], i8** [[G_PRIV_REF]], 82 // LAMBDA: call i32 @__kmpc_reduce( 83 // LAMBDA: switch i32 %{{.+}}, label %[[REDUCTION_DONE:.+]] [ 84 // LAMBDA: i32 1, label %[[CASE1:.+]] 85 // LAMBDA: i32 2, label %[[CASE2:.+]] 86 // LAMBDA: [[CASE1]] 87 // LAMBDA: [[G_VAL:%.+]] = load double, double* [[G]] 88 // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 89 // LAMBDA: [[ADD:%.+]] = fadd double [[G_VAL]], [[G_PRIV_VAL]] 90 // LAMBDA: store double [[ADD]], double* [[G]] 91 // LAMBDA: call void @__kmpc_end_reduce( 92 // LAMBDA: br label %[[REDUCTION_DONE]] 93 // LAMBDA: [[CASE2]] 94 // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 95 // LAMBDA: fadd double 96 // LAMBDA: cmpxchg i64* 97 // LAMBDA: call void @__kmpc_end_reduce( 98 // LAMBDA: br label %[[REDUCTION_DONE]] 99 // LAMBDA: [[REDUCTION_DONE]] 100 // LAMBDA: ret void 101 #pragma omp section 102 [&]() { 103 // LAMBDA: define {{.+}} void [[INNER_LAMBDA]](%{{.+}}* [[ARG_PTR:%.+]]) 104 // LAMBDA: store %{{.+}}* [[ARG_PTR]], %{{.+}}** [[ARG_PTR_REF:%.+]], 105 g = 2; 106 // LAMBDA: [[ARG_PTR:%.+]] = load %{{.+}}*, %{{.+}}** [[ARG_PTR_REF]] 107 // LAMBDA: [[G_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 108 // LAMBDA: [[G_REF:%.+]] = load double*, double** [[G_PTR_REF]] 109 // LAMBDA: store double 2.0{{.+}}, double* [[G_REF]] 110 }(); 111 } 112 }(); 113 return 0; 114 #elif defined(BLOCKS) 115 // BLOCKS: [[G:@.+]] = global double 116 // BLOCKS-LABEL: @main 117 // BLOCKS: call void {{%.+}}(i8 118 ^{ 119 // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8* 120 // BLOCKS: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}) 121 #pragma omp parallel 122 #pragma omp sections reduction(-:g) 123 { 124 // BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}) 125 // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca double, 126 127 // Reduction list for runtime. 128 // BLOCKS: [[RED_LIST:%.+]] = alloca [1 x i8*], 129 130 // BLOCKS: store double 0.0{{.+}}, double* [[G_PRIVATE_ADDR]] 131 g = 1; 132 // BLOCKS: call void @__kmpc_for_static_init_4( 133 // BLOCKS: store double 1.0{{.+}}, double* [[G_PRIVATE_ADDR]], 134 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 135 // BLOCKS: double* [[G_PRIVATE_ADDR]] 136 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 137 // BLOCKS: call void {{%.+}}(i8 138 // BLOCKS: call void @__kmpc_for_static_fini( 139 140 // BLOCKS: [[G_PRIV_REF:%.+]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[RED_LIST]], i64 0, i64 0 141 // BLOCKS: [[BITCAST:%.+]] = bitcast double* [[G_PRIVATE_ADDR]] to i8* 142 // BLOCKS: store i8* [[BITCAST]], i8** [[G_PRIV_REF]], 143 // BLOCKS: call i32 @__kmpc_reduce( 144 // BLOCKS: switch i32 %{{.+}}, label %[[REDUCTION_DONE:.+]] [ 145 // BLOCKS: i32 1, label %[[CASE1:.+]] 146 // BLOCKS: i32 2, label %[[CASE2:.+]] 147 // BLOCKS: [[CASE1]] 148 // BLOCKS: [[G_VAL:%.+]] = load double, double* [[G]] 149 // BLOCKS: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 150 // BLOCKS: [[ADD:%.+]] = fadd double [[G_VAL]], [[G_PRIV_VAL]] 151 // BLOCKS: store double [[ADD]], double* [[G]] 152 // BLOCKS: call void @__kmpc_end_reduce( 153 // BLOCKS: br label %[[REDUCTION_DONE]] 154 // BLOCKS: [[CASE2]] 155 // BLOCKS: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 156 // BLOCKS: fadd double 157 // BLOCKS: cmpxchg i64* 158 // BLOCKS: call void @__kmpc_end_reduce( 159 // BLOCKS: br label %[[REDUCTION_DONE]] 160 // BLOCKS: [[REDUCTION_DONE]] 161 // BLOCKS: ret void 162 #pragma omp section 163 ^{ 164 // BLOCKS: define {{.+}} void {{@.+}}(i8* 165 g = 2; 166 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 167 // BLOCKS: store double 2.0{{.+}}, double* 168 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 169 // BLOCKS: ret 170 }(); 171 } 172 }(); 173 return 0; 174 #else 175 S<float> test; 176 float t_var = 0, t_var1; 177 int vec[] = {1, 2}; 178 S<float> s_arr[] = {1, 2}; 179 S<float> var(3), var1; 180 #pragma omp parallel 181 #pragma omp sections reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) 182 { 183 { 184 vec[0] = t_var; 185 s_arr[0] = var; 186 vec[1] = t_var1; 187 s_arr[1] = var1; 188 } 189 } 190 return tmain<int>(); 191 #endif 192 } 193 194 // CHECK: define {{.*}}i{{[0-9]+}} @main() 195 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]], 196 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]]) 197 // 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 198 // CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT:@.+]]() 199 // CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]* 200 // CHECK: ret 201 // 202 // CHECK: define internal void [[MAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, 203 // CHECK: alloca float, 204 // CHECK: alloca [[S_FLOAT_TY]], 205 // CHECK: alloca [[S_FLOAT_TY]], 206 // CHECK: alloca float, 207 208 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 209 210 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 211 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 212 213 // CHECK-NOT: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 214 // CHECK-NOT: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* 215 216 // CHECK: call void @__kmpc_for_static_init_4( 217 // CHECK: call void @__kmpc_for_static_fini( 218 219 // CHECK: call void @__kmpc_barrier( 220 221 // CHECK: ret void 222 223 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT]]() 224 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]], 225 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]]) 226 // 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 227 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]* 228 // CHECK: ret 229 // 230 // CHECK: define internal void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, 231 // CHECK: alloca i{{[0-9]+}}, 232 // CHECK: alloca i{{[0-9]+}}, 233 // CHECK: alloca i{{[0-9]+}}, 234 // CHECK: alloca i{{[0-9]+}}, 235 // CHECK: alloca i{{[0-9]+}}, 236 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 237 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], 238 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_INT_TY]], 239 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca i{{[0-9]+}}, 240 241 // Reduction list for runtime. 242 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 243 244 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 245 246 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 247 // CHECK: [[VAR_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 248 // CHECK: [[VAR1_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 249 // CHECK: [[T_VAR1_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 250 251 // 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 // For & reduction operation initial value of private variable is ones in all bits. 255 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR_PRIV]]) 256 257 // For && reduction operation initial value of private variable is 1.0. 258 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR1_PRIV]]) 259 260 // For min reduction operation initial value of private variable is largest repesentable value. 261 // CHECK: store i{{[0-9]+}} 2147483647, i{{[0-9]+}}* [[T_VAR1_PRIV]], 262 263 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 264 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 265 // CHECK: call void @__kmpc_for_static_init_4( 266 // Skip checks for internal operations. 267 // CHECK: call void @__kmpc_for_static_fini( 268 269 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 270 271 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 272 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR_PRIV]] to i8* 273 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 274 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 275 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR_PRIV]] to i8* 276 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 277 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 278 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_PRIV]] to i8* 279 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 280 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 281 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR1_PRIV]] to i8* 282 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 283 284 // res = __kmpc_reduce_nowait(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 285 286 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 287 // 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]]) 288 289 // switch(res) 290 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 291 // CHECK: i32 1, label %[[CASE1:.+]] 292 // CHECK: i32 2, label %[[CASE2:.+]] 293 // CHECK: ] 294 295 // case 1: 296 // t_var += t_var_reduction; 297 // CHECK: [[T_VAR_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_REF]], 298 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]], 299 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_VAL]], [[T_VAR_PRIV_VAL]] 300 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_REF]], 301 302 // var = var.operator &(var_reduction); 303 // CHECK: [[UP:%.+]] = call nonnull align 4 dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* nonnull align 4 dereferenceable(4) [[VAR_PRIV]]) 304 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* 305 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 306 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 307 308 // var1 = var1.operator &&(var1_reduction); 309 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) 310 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 311 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 312 // CHECK: [[TRUE]] 313 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) 314 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 315 // CHECK: br label %[[END2]] 316 // CHECK: [[END2]] 317 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 318 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 319 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 320 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* 321 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 322 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 323 324 // t_var1 = min(t_var1, t_var1_reduction); 325 // CHECK: [[T_VAR1_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_REF]], 326 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]], 327 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_VAL]], [[T_VAR1_PRIV_VAL]] 328 // CHECK: br i1 [[CMP]] 329 // CHECK: [[UP:%.+]] = phi i32 330 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_REF]], 331 332 // __kmpc_end_reduce_nowait(<loc>, <gtid>, &<lock>); 333 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 334 335 // break; 336 // CHECK: br label %[[RED_DONE]] 337 338 // case 2: 339 // t_var += t_var_reduction; 340 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]] 341 // CHECK: atomicrmw add i32* [[T_VAR_REF]], i32 [[T_VAR_PRIV_VAL]] monotonic 342 343 // var = var.operator &(var_reduction); 344 // CHECK: call void @__kmpc_critical( 345 // CHECK: [[UP:%.+]] = call nonnull align 4 dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* nonnull align 4 dereferenceable(4) [[VAR_PRIV]]) 346 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* 347 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 348 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 349 // CHECK: call void @__kmpc_end_critical( 350 351 // var1 = var1.operator &&(var1_reduction); 352 // CHECK: call void @__kmpc_critical( 353 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) 354 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 355 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 356 // CHECK: [[TRUE]] 357 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) 358 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 359 // CHECK: br label %[[END2]] 360 // CHECK: [[END2]] 361 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 362 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 363 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 364 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* 365 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 366 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 367 // CHECK: call void @__kmpc_end_critical( 368 369 // t_var1 = min(t_var1, t_var1_reduction); 370 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]] 371 // CHECK: atomicrmw min i32* [[T_VAR1_REF]], i32 [[T_VAR1_PRIV_VAL]] monotonic 372 373 // break; 374 // CHECK: br label %[[RED_DONE]] 375 // CHECK: [[RED_DONE]] 376 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) 377 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* 378 // CHECK: ret void 379 380 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 381 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 382 // ... 383 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 384 // *(Type<n>-1*)rhs[<n>-1]); 385 // } 386 // CHECK: define internal void [[REDUCTION_FUNC]](i8* %0, i8* %1) 387 // t_var_lhs = (i{{[0-9]+}}*)lhs[0]; 388 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 389 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 390 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to i{{[0-9]+}}* 391 // t_var_rhs = (i{{[0-9]+}}*)rhs[0]; 392 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 393 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 394 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to i{{[0-9]+}}* 395 396 // var_lhs = (S<i{{[0-9]+}}>*)lhs[1]; 397 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 398 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 399 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_INT_TY]]* 400 // var_rhs = (S<i{{[0-9]+}}>*)rhs[1]; 401 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 402 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 403 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_INT_TY]]* 404 405 // var1_lhs = (S<i{{[0-9]+}}>*)lhs[2]; 406 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 407 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 408 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_INT_TY]]* 409 // var1_rhs = (S<i{{[0-9]+}}>*)rhs[2]; 410 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 411 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 412 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_INT_TY]]* 413 414 // t_var1_lhs = (i{{[0-9]+}}*)lhs[3]; 415 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 416 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 417 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to i{{[0-9]+}}* 418 // t_var1_rhs = (i{{[0-9]+}}*)rhs[3]; 419 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 420 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 421 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to i{{[0-9]+}}* 422 423 // t_var_lhs += t_var_rhs; 424 // CHECK: [[T_VAR_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_LHS]], 425 // CHECK: [[T_VAR_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_RHS]], 426 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_LHS_VAL]], [[T_VAR_RHS_VAL]] 427 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_LHS]], 428 429 // var_lhs = var_lhs.operator &(var_rhs); 430 // CHECK: [[UP:%.+]] = call nonnull align 4 dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_LHS]], [[S_INT_TY]]* nonnull align 4 dereferenceable(4) [[VAR_RHS]]) 431 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_LHS]] to i8* 432 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 433 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 434 435 // var1_lhs = var1_lhs.operator &&(var1_rhs); 436 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_LHS]]) 437 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 438 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 439 // CHECK: [[TRUE]] 440 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_RHS]]) 441 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 442 // CHECK: br label %[[END2]] 443 // CHECK: [[END2]] 444 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 445 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 446 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 447 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_LHS]] to i8* 448 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 449 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 450 451 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 452 // CHECK: [[T_VAR1_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_LHS]], 453 // CHECK: [[T_VAR1_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_RHS]], 454 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_LHS_VAL]], [[T_VAR1_RHS_VAL]] 455 // CHECK: br i1 [[CMP]] 456 // CHECK: [[UP:%.+]] = phi i32 457 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_LHS]], 458 // CHECK: ret void 459 460 #endif 461 462