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, g_orig; 12 volatile double &g1 = g_orig; 13 14 template <class T> 15 struct S { 16 T f; 17 S(T a) : f(a + g) {} 18 S() : f(g) {} 19 operator T() { return T(); } 20 S &operator&(const S &) { return *this; } 21 ~S() {} 22 }; 23 24 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { float } 25 // CHECK-DAG: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} } 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: [[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 = test; 39 S<T> var1; 40 #pragma omp parallel 41 #pragma omp for reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) nowait 42 for (int i = 0; i < 2; ++i) { 43 vec[i] = t_var; 44 s_arr[i] = var; 45 } 46 #pragma omp parallel 47 #pragma omp for reduction(&& : t_var) 48 for (int i = 0; i < 2; ++i) { 49 vec[i] = t_var; 50 s_arr[i] = var; 51 } 52 return T(); 53 } 54 55 int main() { 56 #ifdef LAMBDA 57 // LAMBDA: [[G:@.+]] = global double 58 // LAMBDA-LABEL: @main 59 // LAMBDA: call void [[OUTER_LAMBDA:@.+]]( 60 [&]() { 61 // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]]( 62 // LAMBDA: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}) 63 #pragma omp parallel 64 #pragma omp for reduction(+:g, g1) 65 for (int i = 0; i < 2; ++i) { 66 // LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}) 67 // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca double, 68 69 // Reduction list for runtime. 70 // LAMBDA: [[RED_LIST:%.+]] = alloca [2 x i8*], 71 72 // LAMBDA: store double 0.0{{.+}}, double* [[G_PRIVATE_ADDR]] 73 // LAMBDA: call void @__kmpc_for_static_init_4( 74 g = 1; 75 g1 = 1; 76 // LAMBDA: store double 1.0{{.+}}, double* [[G_PRIVATE_ADDR]], 77 // LAMBDA: [[G_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 78 // LAMBDA: store double* [[G_PRIVATE_ADDR]], double** [[G_PRIVATE_ADDR_REF]] 79 // LAMBDA: call void [[INNER_LAMBDA:@.+]](%{{.+}}* [[ARG]]) 80 // LAMBDA: call void @__kmpc_for_static_fini( 81 82 // LAMBDA: [[G_PRIV_REF:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[RED_LIST]], i64 0, i64 0 83 // LAMBDA: [[BITCAST:%.+]] = bitcast double* [[G_PRIVATE_ADDR]] to i8* 84 // LAMBDA: store i8* [[BITCAST]], i8** [[G_PRIV_REF]], 85 // LAMBDA: call i32 @__kmpc_reduce( 86 // LAMBDA: switch i32 %{{.+}}, label %[[REDUCTION_DONE:.+]] [ 87 // LAMBDA: i32 1, label %[[CASE1:.+]] 88 // LAMBDA: i32 2, label %[[CASE2:.+]] 89 // LAMBDA: [[CASE1]] 90 // LAMBDA: [[G_VAL:%.+]] = load double, double* [[G]] 91 // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 92 // LAMBDA: [[ADD:%.+]] = fadd double [[G_VAL]], [[G_PRIV_VAL]] 93 // LAMBDA: store double [[ADD]], double* [[G]] 94 // LAMBDA: call void @__kmpc_end_reduce( 95 // LAMBDA: br label %[[REDUCTION_DONE]] 96 // LAMBDA: [[CASE2]] 97 // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 98 // LAMBDA: fadd double 99 // LAMBDA: cmpxchg i64* 100 // LAMBDA: call void @__kmpc_end_reduce( 101 // LAMBDA: br label %[[REDUCTION_DONE]] 102 // LAMBDA: [[REDUCTION_DONE]] 103 // LAMBDA: ret void 104 [&]() { 105 // LAMBDA: define {{.+}} void [[INNER_LAMBDA]](%{{.+}}* [[ARG_PTR:%.+]]) 106 // LAMBDA: store %{{.+}}* [[ARG_PTR]], %{{.+}}** [[ARG_PTR_REF:%.+]], 107 g = 2; 108 g1 = 2; 109 // LAMBDA: [[ARG_PTR:%.+]] = load %{{.+}}*, %{{.+}}** [[ARG_PTR_REF]] 110 // LAMBDA: [[G_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 111 // LAMBDA: [[G_REF:%.+]] = load double*, double** [[G_PTR_REF]] 112 // LAMBDA: store double 2.0{{.+}}, double* [[G_REF]] 113 }(); 114 } 115 }(); 116 return 0; 117 #elif defined(BLOCKS) 118 // BLOCKS: [[G:@.+]] = global double 119 // BLOCKS-LABEL: @main 120 // BLOCKS: call void {{%.+}}(i8 121 ^{ 122 // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8* 123 // BLOCKS: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}) 124 #pragma omp parallel 125 #pragma omp for reduction(-:g, g1) 126 for (int i = 0; i < 2; ++i) { 127 // BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}) 128 // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca double, 129 130 // Reduction list for runtime. 131 // BLOCKS: [[RED_LIST:%.+]] = alloca [2 x i8*], 132 133 // BLOCKS: store double 0.0{{.+}}, double* [[G_PRIVATE_ADDR]] 134 g = 1; 135 g1 = 1; 136 // BLOCKS: call void @__kmpc_for_static_init_4( 137 // BLOCKS: store double 1.0{{.+}}, double* [[G_PRIVATE_ADDR]], 138 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 139 // BLOCKS: double* [[G_PRIVATE_ADDR]] 140 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 141 // BLOCKS: call void {{%.+}}(i8 142 // BLOCKS: call void @__kmpc_for_static_fini( 143 144 // BLOCKS: [[G_PRIV_REF:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[RED_LIST]], i64 0, i64 0 145 // BLOCKS: [[BITCAST:%.+]] = bitcast double* [[G_PRIVATE_ADDR]] to i8* 146 // BLOCKS: store i8* [[BITCAST]], i8** [[G_PRIV_REF]], 147 // BLOCKS: call i32 @__kmpc_reduce( 148 // BLOCKS: switch i32 %{{.+}}, label %[[REDUCTION_DONE:.+]] [ 149 // BLOCKS: i32 1, label %[[CASE1:.+]] 150 // BLOCKS: i32 2, label %[[CASE2:.+]] 151 // BLOCKS: [[CASE1]] 152 // BLOCKS: [[G_VAL:%.+]] = load double, double* [[G]] 153 // BLOCKS: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 154 // BLOCKS: [[ADD:%.+]] = fadd double [[G_VAL]], [[G_PRIV_VAL]] 155 // BLOCKS: store double [[ADD]], double* [[G]] 156 // BLOCKS: call void @__kmpc_end_reduce( 157 // BLOCKS: br label %[[REDUCTION_DONE]] 158 // BLOCKS: [[CASE2]] 159 // BLOCKS: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 160 // BLOCKS: fadd double 161 // BLOCKS: cmpxchg i64* 162 // BLOCKS: call void @__kmpc_end_reduce( 163 // BLOCKS: br label %[[REDUCTION_DONE]] 164 // BLOCKS: [[REDUCTION_DONE]] 165 // BLOCKS: ret void 166 ^{ 167 // BLOCKS: define {{.+}} void {{@.+}}(i8* 168 g = 2; 169 g1 = 2; 170 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 171 // BLOCKS: store double 2.0{{.+}}, double* 172 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 173 // BLOCKS: ret 174 }(); 175 } 176 }(); 177 return 0; 178 #else 179 S<float> test; 180 float t_var = 0, t_var1; 181 int vec[] = {1, 2}; 182 S<float> s_arr[] = {1, 2}; 183 S<float> &var = test; 184 S<float> var1, arrs[10][4]; 185 #pragma omp parallel 186 #pragma omp for reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) 187 for (int i = 0; i < 2; ++i) { 188 vec[i] = t_var; 189 s_arr[i] = var; 190 } 191 int arr[10][vec[1]]; 192 #pragma omp parallel for reduction(+:arr[1][:vec[1]]) reduction(&:arrs[1:vec[1]][1:2]) 193 for (int i = 0; i < 10; ++i) 194 ++arr[1][i]; 195 return tmain<int>(); 196 #endif 197 } 198 199 // CHECK: define {{.*}}i{{[0-9]+}} @main() 200 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]], 201 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]]) 202 // 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 203 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 5, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, i64, i64, i32*, [2 x i32]*, [10 x [4 x [[S_FLOAT_TY]]]]*)* [[MAIN_MICROTASK1:@.+]] to void 204 // CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT:@.+]]() 205 // CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]* 206 // CHECK: ret 207 // 208 // CHECK: define internal void [[MAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, float* dereferenceable(4) %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}, float* dereferenceable(4) %{{.+}}, [2 x i32]* dereferenceable(8) %vec, [2 x [[S_FLOAT_TY]]]* dereferenceable(8) %{{.+}}) 209 // CHECK: [[T_VAR_PRIV:%.+]] = alloca float, 210 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 211 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 212 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca float, 213 214 // Reduction list for runtime. 215 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 216 217 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 218 219 // CHECK: [[T_VAR_REF:%.+]] = load float*, float** % 220 // CHECK: [[VAR1_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** % 221 // CHECK: [[T_VAR1_REF:%.+]] = load float*, float** % 222 223 // For + reduction operation initial value of private variable is 0. 224 // CHECK: store float 0.0{{.+}}, float* [[T_VAR_PRIV]], 225 226 // For & reduction operation initial value of private variable is ones in all bits. 227 // CHECK: [[VAR_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** % 228 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 229 230 // For && reduction operation initial value of private variable is 1.0. 231 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[VAR1_PRIV]]) 232 233 // For min reduction operation initial value of private variable is largest repesentable value. 234 // CHECK: store float 0x47EFFFFFE0000000, float* [[T_VAR1_PRIV]], 235 236 237 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 238 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 239 // CHECK: call void @__kmpc_for_static_init_4( 240 // Skip checks for internal operations. 241 // CHECK: call void @__kmpc_for_static_fini( 242 243 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 244 245 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 246 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR_PRIV]] to i8* 247 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 248 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 249 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_PRIV]] to i8* 250 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 251 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 252 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_PRIV]] to i8* 253 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 254 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 255 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR1_PRIV]] to i8* 256 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 257 258 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 259 260 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 261 // CHECK: [[RES:%.+]] = call i32 @__kmpc_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], i32 4, i64 32, i8* [[BITCAST]], void (i8*, i8*)* [[REDUCTION_FUNC:@.+]], [8 x i32]* [[REDUCTION_LOCK]]) 262 263 // switch(res) 264 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 265 // CHECK: i32 1, label %[[CASE1:.+]] 266 // CHECK: i32 2, label %[[CASE2:.+]] 267 // CHECK: ] 268 269 // case 1: 270 // t_var += t_var_reduction; 271 // CHECK: [[T_VAR_VAL:%.+]] = load float, float* [[T_VAR_REF]], 272 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load float, float* [[T_VAR_PRIV]], 273 // CHECK: [[UP:%.+]] = fadd float [[T_VAR_VAL]], [[T_VAR_PRIV_VAL]] 274 // CHECK: store float [[UP]], float* [[T_VAR_REF]], 275 276 // var = var.operator &(var_reduction); 277 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @{{.+}}([[S_FLOAT_TY]]* [[VAR_REF]], [[S_FLOAT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 278 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_REF]] to i8* 279 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[UP]] to i8* 280 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 281 282 // var1 = var1.operator &&(var1_reduction); 283 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_REF]]) 284 // CHECK: [[VAR1_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 285 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 286 // CHECK: [[TRUE]] 287 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_PRIV]]) 288 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 289 // CHECK: br label %[[END2]] 290 // CHECK: [[END2]] 291 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 292 // CHECK: [[CONV:%.+]] = uitofp i1 [[COND_LVALUE]] to float 293 // CHECK: call void @{{.+}}([[S_FLOAT_TY]]* [[COND_LVALUE:%.+]], float [[CONV]]) 294 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_REF]] to i8* 295 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[COND_LVALUE]] to i8* 296 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 297 298 // t_var1 = min(t_var1, t_var1_reduction); 299 // CHECK: [[T_VAR1_VAL:%.+]] = load float, float* [[T_VAR1_REF]], 300 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load float, float* [[T_VAR1_PRIV]], 301 // CHECK: [[CMP:%.+]] = fcmp olt float [[T_VAR1_VAL]], [[T_VAR1_PRIV_VAL]] 302 // CHECK: br i1 [[CMP]] 303 // CHECK: [[UP:%.+]] = phi float 304 // CHECK: store float [[UP]], float* [[T_VAR1_REF]], 305 306 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 307 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 308 309 // break; 310 // CHECK: br label %[[RED_DONE]] 311 312 // case 2: 313 // t_var += t_var_reduction; 314 // CHECK: load float, float* [[T_VAR_PRIV]] 315 // CHECK: [[T_VAR_REF_INT:%.+]] = bitcast float* [[T_VAR_REF]] to i32* 316 // CHECK: [[OLD1:%.+]] = load atomic i32, i32* [[T_VAR_REF_INT]] monotonic, 317 // CHECK: br label %[[CONT:.+]] 318 // CHECK: [[CONT]] 319 // CHECK: [[ORIG_OLD_INT:%.+]] = phi i32 [ [[OLD1]], %{{.+}} ], [ [[OLD2:%.+]], %[[CONT]] ] 320 // CHECK: fadd float 321 // CHECK: [[UP_INT:%.+]] = load i32, i32* 322 // CHECK: [[T_VAR_REF_INT:%.+]] = bitcast float* [[T_VAR_REF]] to i32* 323 // CHECK: [[RES:%.+]] = cmpxchg i32* [[T_VAR_REF_INT]], i32 [[ORIG_OLD_INT]], i32 [[UP_INT]] monotonic monotonic 324 // CHECK: [[OLD2:%.+]] = extractvalue { i32, i1 } [[RES]], 0 325 // CHECK: [[SUCCESS_FAIL:%.+]] = extractvalue { i32, i1 } [[RES]], 1 326 // CHECK: br i1 [[SUCCESS_FAIL]], label %[[ATOMIC_DONE:.+]], label %[[CONT]] 327 // CHECK: [[ATOMIC_DONE]] 328 329 // var = var.operator &(var_reduction); 330 // CHECK: call void @__kmpc_critical( 331 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @{{.+}}([[S_FLOAT_TY]]* [[VAR_REF]], [[S_FLOAT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 332 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_REF]] to i8* 333 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[UP]] to i8* 334 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 335 // CHECK: call void @__kmpc_end_critical( 336 337 // var1 = var1.operator &&(var1_reduction); 338 // CHECK: call void @__kmpc_critical( 339 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_REF]]) 340 // CHECK: [[VAR1_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 341 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 342 // CHECK: [[TRUE]] 343 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_PRIV]]) 344 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 345 // CHECK: br label %[[END2]] 346 // CHECK: [[END2]] 347 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 348 // CHECK: [[CONV:%.+]] = uitofp i1 [[COND_LVALUE]] to float 349 // CHECK: call void @{{.+}}([[S_FLOAT_TY]]* [[COND_LVALUE:%.+]], float [[CONV]]) 350 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_REF]] to i8* 351 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[COND_LVALUE]] to i8* 352 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 353 // CHECK: call void @__kmpc_end_critical( 354 355 // t_var1 = min(t_var1, t_var1_reduction); 356 // CHECK: load float, float* [[T_VAR1_PRIV]] 357 // CHECK: [[T_VAR1_REF_INT:%.+]] = bitcast float* [[T_VAR1_REF]] to i32* 358 // CHECK: [[OLD1:%.+]] = load atomic i32, i32* [[T_VAR1_REF_INT]] monotonic, 359 // CHECK: br label %[[CONT:.+]] 360 // CHECK: [[CONT]] 361 // CHECK: [[ORIG_OLD_INT:%.+]] = phi i32 [ [[OLD1]], %{{.+}} ], [ [[OLD2:%.+]], %{{.+}} ] 362 // CHECK: [[CMP:%.+]] = fcmp olt float 363 // CHECK: br i1 [[CMP]] 364 // CHECK: phi float 365 // CHECK: [[UP_INT:%.+]] = load i32 366 // CHECK: [[T_VAR1_REF_INT:%.+]] = bitcast float* [[T_VAR1_REF]] to i32* 367 // CHECK: [[RES:%.+]] = cmpxchg i32* [[T_VAR1_REF_INT]], i32 [[ORIG_OLD_INT]], i32 [[UP_INT]] monotonic monotonic 368 // CHECK: [[OLD2:%.+]] = extractvalue { i32, i1 } [[RES]], 0 369 // CHECK: [[SUCCESS_FAIL:%.+]] = extractvalue { i32, i1 } [[RES]], 1 370 // CHECK: br i1 [[SUCCESS_FAIL]], label %[[ATOMIC_DONE:.+]], label %[[CONT]] 371 // CHECK: [[ATOMIC_DONE]] 372 373 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 374 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 375 376 // break; 377 // CHECK: br label %[[RED_DONE]] 378 // CHECK: [[RED_DONE]] 379 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 380 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* 381 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 382 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 383 384 // CHECK: ret void 385 386 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 387 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 388 // ... 389 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 390 // *(Type<n>-1*)rhs[<n>-1]); 391 // } 392 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 393 // t_var_lhs = (float*)lhs[0]; 394 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 395 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 396 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to float* 397 // t_var_rhs = (float*)rhs[0]; 398 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 399 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 400 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to float* 401 402 // var_lhs = (S<float>*)lhs[1]; 403 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 404 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 405 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_FLOAT_TY]]* 406 // var_rhs = (S<float>*)rhs[1]; 407 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 408 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 409 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_FLOAT_TY]]* 410 411 // var1_lhs = (S<float>*)lhs[2]; 412 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 413 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 414 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_FLOAT_TY]]* 415 // var1_rhs = (S<float>*)rhs[2]; 416 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 417 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 418 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_FLOAT_TY]]* 419 420 // t_var1_lhs = (float*)lhs[3]; 421 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 422 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 423 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to float* 424 // t_var1_rhs = (float*)rhs[3]; 425 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 426 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 427 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to float* 428 429 // t_var_lhs += t_var_rhs; 430 // CHECK: [[T_VAR_LHS_VAL:%.+]] = load float, float* [[T_VAR_LHS]], 431 // CHECK: [[T_VAR_RHS_VAL:%.+]] = load float, float* [[T_VAR_RHS]], 432 // CHECK: [[UP:%.+]] = fadd float [[T_VAR_LHS_VAL]], [[T_VAR_RHS_VAL]] 433 // CHECK: store float [[UP]], float* [[T_VAR_LHS]], 434 435 // var_lhs = var_lhs.operator &(var_rhs); 436 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @{{.+}}([[S_FLOAT_TY]]* [[VAR_LHS]], [[S_FLOAT_TY]]* dereferenceable(4) [[VAR_RHS]]) 437 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_LHS]] to i8* 438 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[UP]] to i8* 439 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 440 441 // var1_lhs = var1_lhs.operator &&(var1_rhs); 442 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_LHS]]) 443 // CHECK: [[VAR1_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 444 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 445 // CHECK: [[TRUE]] 446 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_RHS]]) 447 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 448 // CHECK: br label %[[END2]] 449 // CHECK: [[END2]] 450 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 451 // CHECK: [[CONV:%.+]] = uitofp i1 [[COND_LVALUE]] to float 452 // CHECK: call void @{{.+}}([[S_FLOAT_TY]]* [[COND_LVALUE:%.+]], float [[CONV]]) 453 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_LHS]] to i8* 454 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[COND_LVALUE]] to i8* 455 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 456 457 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 458 // CHECK: [[T_VAR1_LHS_VAL:%.+]] = load float, float* [[T_VAR1_LHS]], 459 // CHECK: [[T_VAR1_RHS_VAL:%.+]] = load float, float* [[T_VAR1_RHS]], 460 // CHECK: [[CMP:%.+]] = fcmp olt float [[T_VAR1_LHS_VAL]], [[T_VAR1_RHS_VAL]] 461 // CHECK: br i1 [[CMP]] 462 // CHECK: [[UP:%.+]] = phi float 463 // CHECK: store float [[UP]], float* [[T_VAR1_LHS]], 464 // CHECK: ret void 465 466 // CHECK: define internal void [[MAIN_MICROTASK1]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i64 %{{.+}}, i64 %{{.+}}, i32* nonnull %{{.+}}, [2 x i32]* dereferenceable(8) %{{.+}}, [10 x [4 x [[S_FLOAT_TY]]]]* dereferenceable(160) %{{.+}}) 467 468 // Reduction list for runtime. 469 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 470 471 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 472 473 // CHECK: [[IDX1:%.+]] = mul nsw i64 1, %{{.+}} 474 // CHECK: [[LB1:%.+]] = getelementptr inbounds i32, i32* %{{.+}}, i64 [[IDX1]] 475 // CHECK: [[LB1_0:%.+]] = getelementptr inbounds i32, i32* [[LB1]], i64 0 476 // CHECK: [[IDX1:%.+]] = mul nsw i64 1, %{{.+}} 477 // CHECK: [[UB1:%.+]] = getelementptr inbounds i32, i32* %{{.+}}, i64 [[IDX1]] 478 // CHECK: [[UB1_UP:%.+]] = getelementptr inbounds i32, i32* [[UB1]], i64 % 479 // CHECK: [[UB_CAST:%.+]] = ptrtoint i32* [[UB1_UP]] to i64 480 // CHECK: [[LB_CAST:%.+]] = ptrtoint i32* [[LB1_0]] to i64 481 // CHECK: [[DIFF:%.+]] = sub i64 [[UB_CAST]], [[LB_CAST]] 482 // CHECK: [[SIZE_1:%.+]] = sdiv exact i64 [[DIFF]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64) 483 // CHECK: [[ARR_SIZE:%.+]] = add nuw i64 [[SIZE_1]], 1 484 // CHECK: call i8* @llvm.stacksave() 485 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]], 486 487 // Check initialization of private copy. 488 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]] 489 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]] 490 // CHECK: br i1 [[ISEMPTY]], 491 // CHECK: phi i32* 492 // CHECK: store i32 0, i32* % 493 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 494 // CHECK: br i1 [[DONE]], 495 496 // CHECK: [[ARRS_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[ARRS_SIZE:%.+]], 497 498 // Check initialization of private copy. 499 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]] 500 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]] 501 // CHECK: br i1 [[ISEMPTY]], 502 // CHECK: phi [[S_FLOAT_TY]]* 503 // CHECK: call void @_ZN1SIfEC1Ev([[S_FLOAT_TY]]* % 504 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 505 // CHECK: br i1 [[DONE]], 506 507 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 508 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 509 // CHECK: call void @__kmpc_for_static_init_4( 510 // Skip checks for internal operations. 511 // CHECK: call void @__kmpc_for_static_fini( 512 513 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 514 515 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 516 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8* 517 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]], 518 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 519 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8* 520 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]], 521 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 522 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV]] to i8* 523 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]], 524 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 525 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARRS_SIZE]] to i8* 526 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_SIZE_REF]], 527 528 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 529 530 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 531 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 532 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 533 // CHECK: [[RES:%.+]] = call i32 @__kmpc_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], i32 2, i64 32, i8* [[BITCAST]], void (i8*, i8*)* [[REDUCTION_FUNC:@.+]], [8 x i32]* [[REDUCTION_LOCK]]) 534 535 // switch(res) 536 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 537 // CHECK: i32 1, label %[[CASE1:.+]] 538 // CHECK: i32 2, label %[[CASE2:.+]] 539 // CHECK: ] 540 541 // case 1: 542 // CHECK: [[CASE1]] 543 544 // arr[:] += arr_reduction[:]; 545 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 546 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 547 // CHECK: br i1 [[ISEMPTY]], 548 // CHECK: phi i32* 549 // CHECK: [[ADD:%.+]] = add nsw i32 % 550 // CHECK: store i32 [[ADD]], i32* % 551 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 552 // CHECK: br i1 [[DONE]], 553 554 // arrs[:] = var.operator &(arrs_reduction[:]); 555 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 556 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 557 // CHECK: br i1 [[ISEMPTY]], 558 // CHECK: phi [[S_FLOAT_TY]]* 559 // CHECK: [[AND:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}) 560 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 561 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %{{.+}}, i8* [[BITCAST]], i64 4, i32 4, i1 false) 562 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 563 // CHECK: br i1 [[DONE]], 564 565 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 566 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 567 568 // break; 569 // CHECK: br label %[[RED_DONE]] 570 571 // case 2: 572 // CHECK: [[CASE2]] 573 574 // arr[:] += arr_reduction[:]; 575 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 576 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 577 // CHECK: br i1 [[ISEMPTY]], 578 // CHECK: phi i32* 579 // CHECK: atomicrmw add i32* %{{.+}}, i32 %{{.+}} monotonic 580 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 581 // CHECK: br i1 [[DONE]], 582 583 // arrs[:] = var.operator &(arrs_reduction[:]); 584 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 585 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 586 // CHECK: br i1 [[ISEMPTY]], 587 // CHECK: phi [[S_FLOAT_TY]]* 588 // CHECK: call void @__kmpc_critical( 589 // CHECK: [[AND:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}) 590 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 591 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %{{.+}}, i8* [[BITCAST]], i64 4, i32 4, i1 false) 592 // CHECK: call void @__kmpc_end_critical( 593 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 594 // CHECK: br i1 [[DONE]], 595 596 // break; 597 // CHECK: br label %[[RED_DONE]] 598 // CHECK: [[RED_DONE]] 599 600 // Check destruction of private copy. 601 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]] 602 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]] 603 // CHECK: br i1 [[ISEMPTY]], 604 // CHECK: phi [[S_FLOAT_TY]]* 605 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* % 606 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[ARRS_PRIV]] 607 // CHECK: br i1 [[DONE]], 608 // CHECK: call void @llvm.stackrestore(i8* 609 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 610 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 611 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 612 613 // CHECK: ret void 614 615 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 616 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 617 // ... 618 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 619 // *(Type<n>-1*)rhs[<n>-1]); 620 // } 621 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 622 // arr_rhs = (int*)rhs[0]; 623 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 624 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]], 625 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32* 626 // arr_lhs = (int*)lhs[0]; 627 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 628 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]], 629 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32* 630 631 // arr_size = (size_t)lhs[1]; 632 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 633 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]], 634 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64 635 636 // arrs_rhs = (S<float>*)rhs[2]; 637 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 638 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]], 639 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]* 640 // arrs_lhs = (S<float>*)lhs[2]; 641 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 642 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]], 643 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]* 644 645 // arrs_size = (size_t)lhs[3]; 646 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 647 // CHECK: [[ARRS_SIZE_VOID:%.+]] = load i8*, i8** [[ARRS_SIZE_REF]], 648 // CHECK: [[ARRS_SIZE:%.+]] = ptrtoint i8* [[ARRS_SIZE_VOID]] to i64 649 650 // arr_lhs[:] += arr_rhs[:]; 651 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]] 652 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]] 653 // CHECK: br i1 [[ISEMPTY]], 654 // CHECK: phi i32* 655 // CHECK: [[ADD:%.+]] = add nsw i32 % 656 // CHECK: store i32 [[ADD]], i32* % 657 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 658 // CHECK: br i1 [[DONE]], 659 660 // arrs_lhs = arrs_lhs.operator &(arrs_rhs); 661 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 662 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 663 // CHECK: br i1 [[ISEMPTY]], 664 // CHECK: phi [[S_FLOAT_TY]]* 665 // CHECK: [[AND:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}) 666 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 667 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %{{.+}}, i8* [[BITCAST]], i64 4, i32 4, i1 false) 668 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 669 // CHECK: br i1 [[DONE]], 670 671 // CHECK: ret void 672 673 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT]]() 674 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]], 675 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]]) 676 // 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 677 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]* 678 // CHECK: ret 679 // 680 // CHECK: define internal void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i32* dereferenceable(4) %{{.+}}, [[S_INT_TY]]* dereferenceable(4) %{{.+}}, [[S_INT_TY]]* dereferenceable(4) %{{.+}}, i32* dereferenceable(4) %{{.+}}, [2 x i32]* dereferenceable(8) %{{.+}}, [2 x [[S_INT_TY]]]* dereferenceable(8) %{{.+}}) 681 // CHECK: alloca i{{[0-9]+}}, 682 // CHECK: alloca i{{[0-9]+}}, 683 // CHECK: alloca i{{[0-9]+}}, 684 // CHECK: alloca i{{[0-9]+}}, 685 // CHECK: alloca i{{[0-9]+}}, 686 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 687 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], 688 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_INT_TY]], 689 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca i{{[0-9]+}}, 690 691 // Reduction list for runtime. 692 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 693 694 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 695 696 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 697 // CHECK: [[VAR1_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 698 // CHECK: [[T_VAR1_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 699 700 // For + reduction operation initial value of private variable is 0. 701 // CHECK: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[T_VAR_PRIV]], 702 703 // For & reduction operation initial value of private variable is ones in all bits. 704 // CHECK: [[VAR_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 705 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR_PRIV]]) 706 707 // For && reduction operation initial value of private variable is 1.0. 708 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR1_PRIV]]) 709 710 // For min reduction operation initial value of private variable is largest repesentable value. 711 // CHECK: store i{{[0-9]+}} 2147483647, i{{[0-9]+}}* [[T_VAR1_PRIV]], 712 713 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 714 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 715 // CHECK: call void @__kmpc_for_static_init_4( 716 // Skip checks for internal operations. 717 // CHECK: call void @__kmpc_for_static_fini( 718 719 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 720 721 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 722 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR_PRIV]] to i8* 723 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 724 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 725 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR_PRIV]] to i8* 726 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 727 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 728 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_PRIV]] to i8* 729 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 730 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 731 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR1_PRIV]] to i8* 732 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 733 734 // res = __kmpc_reduce_nowait(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 735 736 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 737 // 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]]) 738 739 // switch(res) 740 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 741 // CHECK: i32 1, label %[[CASE1:.+]] 742 // CHECK: i32 2, label %[[CASE2:.+]] 743 // CHECK: ] 744 745 // case 1: 746 // t_var += t_var_reduction; 747 // CHECK: [[T_VAR_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_REF]], 748 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]], 749 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_VAL]], [[T_VAR_PRIV_VAL]] 750 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_REF]], 751 752 // var = var.operator &(var_reduction); 753 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 754 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* 755 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 756 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 757 758 // var1 = var1.operator &&(var1_reduction); 759 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) 760 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 761 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 762 // CHECK: [[TRUE]] 763 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) 764 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 765 // CHECK: br label %[[END2]] 766 // CHECK: [[END2]] 767 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 768 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 769 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 770 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* 771 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 772 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 773 774 // t_var1 = min(t_var1, t_var1_reduction); 775 // CHECK: [[T_VAR1_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_REF]], 776 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]], 777 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_VAL]], [[T_VAR1_PRIV_VAL]] 778 // CHECK: br i1 [[CMP]] 779 // CHECK: [[UP:%.+]] = phi i32 780 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_REF]], 781 782 // __kmpc_end_reduce_nowait(<loc>, <gtid>, &<lock>); 783 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 784 785 // break; 786 // CHECK: br label %[[RED_DONE]] 787 788 // case 2: 789 // t_var += t_var_reduction; 790 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]] 791 // CHECK: atomicrmw add i32* [[T_VAR_REF]], i32 [[T_VAR_PRIV_VAL]] monotonic 792 793 // var = var.operator &(var_reduction); 794 // CHECK: call void @__kmpc_critical( 795 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 796 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* 797 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 798 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 799 // CHECK: call void @__kmpc_end_critical( 800 801 // var1 = var1.operator &&(var1_reduction); 802 // CHECK: call void @__kmpc_critical( 803 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) 804 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 805 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 806 // CHECK: [[TRUE]] 807 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) 808 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 809 // CHECK: br label %[[END2]] 810 // CHECK: [[END2]] 811 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 812 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 813 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 814 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* 815 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 816 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 817 // CHECK: call void @__kmpc_end_critical( 818 819 // t_var1 = min(t_var1, t_var1_reduction); 820 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]] 821 // CHECK: atomicrmw min i32* [[T_VAR1_REF]], i32 [[T_VAR1_PRIV_VAL]] monotonic 822 823 // break; 824 // CHECK: br label %[[RED_DONE]] 825 // CHECK: [[RED_DONE]] 826 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) 827 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* 828 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 829 // CHECK: ret void 830 831 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 832 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 833 // ... 834 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 835 // *(Type<n>-1*)rhs[<n>-1]); 836 // } 837 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 838 // t_var_lhs = (i{{[0-9]+}}*)lhs[0]; 839 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 840 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 841 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to i{{[0-9]+}}* 842 // t_var_rhs = (i{{[0-9]+}}*)rhs[0]; 843 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 844 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 845 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to i{{[0-9]+}}* 846 847 // var_lhs = (S<i{{[0-9]+}}>*)lhs[1]; 848 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 849 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 850 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_INT_TY]]* 851 // var_rhs = (S<i{{[0-9]+}}>*)rhs[1]; 852 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 853 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 854 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_INT_TY]]* 855 856 // var1_lhs = (S<i{{[0-9]+}}>*)lhs[2]; 857 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 858 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 859 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_INT_TY]]* 860 // var1_rhs = (S<i{{[0-9]+}}>*)rhs[2]; 861 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 862 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 863 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_INT_TY]]* 864 865 // t_var1_lhs = (i{{[0-9]+}}*)lhs[3]; 866 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 867 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 868 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to i{{[0-9]+}}* 869 // t_var1_rhs = (i{{[0-9]+}}*)rhs[3]; 870 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 871 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 872 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to i{{[0-9]+}}* 873 874 // t_var_lhs += t_var_rhs; 875 // CHECK: [[T_VAR_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_LHS]], 876 // CHECK: [[T_VAR_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_RHS]], 877 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_LHS_VAL]], [[T_VAR_RHS_VAL]] 878 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_LHS]], 879 880 // var_lhs = var_lhs.operator &(var_rhs); 881 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_LHS]], [[S_INT_TY]]* dereferenceable(4) [[VAR_RHS]]) 882 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_LHS]] to i8* 883 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 884 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 885 886 // var1_lhs = var1_lhs.operator &&(var1_rhs); 887 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_LHS]]) 888 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 889 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 890 // CHECK: [[TRUE]] 891 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_RHS]]) 892 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 893 // CHECK: br label %[[END2]] 894 // CHECK: [[END2]] 895 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 896 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 897 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 898 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_LHS]] to i8* 899 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 900 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 901 902 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 903 // CHECK: [[T_VAR1_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_LHS]], 904 // CHECK: [[T_VAR1_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_RHS]], 905 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_LHS_VAL]], [[T_VAR1_RHS_VAL]] 906 // CHECK: br i1 [[CMP]] 907 // CHECK: [[UP:%.+]] = phi i32 908 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_LHS]], 909 // CHECK: ret void 910 911 #endif 912 913