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 // expected-no-diagnostics 5 #ifndef HEADER 6 #define HEADER 7 8 volatile double g, g_orig; 9 volatile double &g1 = g_orig; 10 11 struct BaseS { 12 int x; 13 }; 14 struct BaseS1 { 15 float y; 16 }; 17 18 template <class T> 19 struct S : public BaseS, public BaseS1 { 20 T f; 21 S(T a) : f(a + g) {} 22 S() : f(g) {} 23 ~S() {} 24 }; 25 void red(BaseS1&, const BaseS1&); 26 void red_plus(BaseS1&, const BaseS1&); 27 void init(BaseS1&, const BaseS1&); 28 void init1(BaseS1&, const BaseS1&); 29 void init2(BaseS1&, const BaseS1&); 30 void init_plus(BaseS1&, const BaseS1&); 31 #pragma omp declare reduction(operator& : BaseS1 : red(omp_out, omp_in)) initializer(init(omp_priv, omp_orig)) 32 #pragma omp declare reduction(+ : BaseS1 : red_plus(omp_out, omp_in)) initializer(init_plus(omp_priv, omp_orig)) 33 #pragma omp declare reduction(&& : S<float>, S<int> : omp_out.f *= omp_in.f) initializer(init1(omp_priv, omp_orig)) 34 35 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { %{{[^,]+}}, %{{[^,]+}}, float } 36 // CHECK-DAG: [[S_INT_TY:%.+]] = type { %{{[^,]+}}, %{{[^,]+}}, i{{[0-9]+}} } 37 // CHECK-DAG: [[ATOMIC_REDUCE_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 38 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8* 39 // CHECK-DAG: [[REDUCTION_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 40 // CHECK-DAG: [[REDUCTION_LOCK:@.+]] = common global [8 x i32] zeroinitializer 41 42 #pragma omp declare reduction(operator&& : int : omp_out = 111 & omp_in) 43 template <typename T> 44 T tmain() { 45 T t; 46 S<T> test; 47 T t_var = T(), t_var1; 48 T vec[] = {1, 2}; 49 S<T> s_arr[] = {1, 2}; 50 S<T> &var = test; 51 S<T> var1; 52 #pragma omp declare reduction(operator& : T : omp_out = 15 + omp_in) 53 #pragma omp declare reduction(operator+ : T : omp_out = 1513 + omp_in) initializer(omp_priv = 321) 54 #pragma omp declare reduction(min : T : omp_out = 47 - omp_in) initializer(omp_priv = 432 / omp_orig) 55 #pragma omp declare reduction(operator&& : S<T> : omp_out.f = 17 * omp_in.f) initializer(init2(omp_priv, omp_orig)) 56 #pragma omp declare reduction(operator&& : T : omp_out = 17 * omp_in) 57 #pragma omp parallel 58 #pragma omp for reduction(+ : t_var) reduction(& : var) reduction(&& : var1) reduction(min : t_var1) nowait 59 for (int i = 0; i < 2; ++i) { 60 vec[i] = t_var; 61 s_arr[i] = var; 62 } 63 #pragma omp parallel 64 #pragma omp for reduction(&& : t_var) 65 for (int i = 0; i < 2; ++i) { 66 vec[i] = t_var; 67 s_arr[i] = var; 68 } 69 return T(); 70 } 71 72 extern S<float> **foo(); 73 74 #pragma omp declare reduction(operator- : float, double : omp_out = 333 + omp_in) 75 #pragma omp declare reduction(min : float, double : omp_out = 555 + omp_in) 76 int main() { 77 #pragma omp declare reduction(operator+ : float, double : omp_out = 222 - omp_in) initializer(omp_priv = -1) 78 S<float> test; 79 float t_var = 0, t_var1; 80 int vec[] = {1, 2}; 81 S<float> s_arr[] = {1, 2}; 82 S<float> &var = test; 83 S<float> var1, arrs[10][4]; 84 S<float> **var2 = foo(); 85 S<float> vvar2[2]; 86 S<float>(&var3)[2] = s_arr; 87 #pragma omp declare reduction(operator+ : int : omp_out = 555 * omp_in) initializer(omp_priv = 888) 88 #pragma omp parallel 89 #pragma omp for reduction(+ : t_var) reduction(& : var) reduction(&& : var1) reduction(min : t_var1) 90 for (int i = 0; i < 2; ++i) { 91 vec[i] = t_var; 92 s_arr[i] = var; 93 } 94 int arr[10][vec[1]]; 95 #pragma omp parallel for reduction(+ : arr[1][ : vec[1]]) reduction(& : arrs[1 : vec[1]][1 : 2]) 96 for (int i = 0; i < 10; ++i) 97 ++arr[1][i]; 98 #pragma omp parallel 99 #pragma omp for reduction(+ : arr) reduction(& : arrs) 100 for (int i = 0; i < 10; ++i) 101 ++arr[1][i]; 102 #pragma omp parallel 103 #pragma omp for reduction(& : var2[0 : 5][1 : 6]) 104 for (int i = 0; i < 10; ++i) 105 ; 106 #pragma omp parallel 107 #pragma omp for reduction(& : vvar2[0 : 5]) 108 for (int i = 0; i < 10; ++i) 109 ; 110 #pragma omp parallel 111 #pragma omp for reduction(& : var3[1 : 2]) 112 for (int i = 0; i < 10; ++i) 113 ; 114 #pragma omp parallel 115 #pragma omp for reduction(& : var3) 116 for (int i = 0; i < 10; ++i) 117 ; 118 return tmain<int>(); 119 } 120 121 // CHECK: define {{.*}}i{{[0-9]+}} @main() 122 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]], 123 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]]) 124 // 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 125 // 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 126 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, i64, i64, i32*, [10 x [4 x [[S_FLOAT_TY]]]]*)* [[MAIN_MICROTASK2:@.+]] to void 127 // 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]+}}*, [[S_FLOAT_TY]]***)* [[MAIN_MICROTASK3:@.+]] to void 128 // 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]+}}*, [2 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK4:@.+]] to void 129 // 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]+}}*, [2 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK5:@.+]] to void 130 // 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]+}}*, [2 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK6:@.+]] to void 131 // CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT:@.+]]() 132 // CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]* 133 // CHECK: ret 134 // 135 // CHECK: define internal void [[MAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, float* dereferenceable(4) %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(12) %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(12) %{{.+}}, float* dereferenceable(4) %{{.+}}, [2 x i32]* dereferenceable(8) %vec, [2 x [[S_FLOAT_TY]]]* dereferenceable(24) %{{.+}}) 136 // CHECK: [[T_VAR_PRIV:%.+]] = alloca float, 137 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 138 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 139 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca float, 140 141 // Reduction list for runtime. 142 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 143 144 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 145 146 // CHECK: [[T_VAR_REF:%.+]] = load float*, float** % 147 // CHECK: [[VAR1_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** % 148 // CHECK: [[T_VAR1_REF:%.+]] = load float*, float** % 149 150 // For + reduction operation initial value of private variable is -1. 151 // CHECK: store float -1.0{{.+}}, float* 152 153 // For & reduction operation initial value of private variable is defined by call of 'init()' function. 154 // CHECK: call {{.*}}void @_Z4initR6BaseS1RKS_( 155 156 // For && reduction operation initial value of private variable is 1.0. 157 // CHECK: call {{.*}}void @_Z5init1R6BaseS1RKS_( 158 159 // For min reduction operation initial value of private variable is largest repesentable value. 160 // CHECK: [[INIT:%.+]] = load float, float* @ 161 // CHECK: store float [[INIT]], float* [[T_VAR1_PRIV]], 162 163 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 164 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 165 // CHECK: call void @__kmpc_for_static_init_4( 166 // Skip checks for internal operations. 167 // CHECK: call void @__kmpc_for_static_fini( 168 169 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 170 171 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 172 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR_PRIV]] to i8* 173 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 174 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 175 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_PRIV]] to i8* 176 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 177 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 178 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_PRIV]] to i8* 179 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 180 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 181 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR1_PRIV]] to i8* 182 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 183 184 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 185 186 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 187 // 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]]) 188 189 // switch(res) 190 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 191 // CHECK: i32 1, label %[[CASE1:.+]] 192 // CHECK: i32 2, label %[[CASE2:.+]] 193 // CHECK: ] 194 195 // case 1: 196 // t_var += t_var_reduction; 197 // CHECK: fsub float 2.220000e+02, % 198 199 // var = var.operator &(var_reduction); 200 // CHECK: call {{.*}}void @_Z3redR6BaseS1RKS_( 201 202 // var1 = var1.operator &&(var1_reduction); 203 // CHECK: fmul float 204 205 // t_var1 = min(t_var1, t_var1_reduction); 206 // CHECK: fadd float 5.550000e+02, % 207 208 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 209 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 210 211 // break; 212 // CHECK: br label %[[RED_DONE]] 213 214 // case 2: 215 // t_var += t_var_reduction; 216 // CHECK: call void @__kmpc_critical( 217 // CHECK: fsub float 2.220000e+02, % 218 // CHECK: call void @__kmpc_end_critical( 219 220 // var = var.operator &(var_reduction); 221 // CHECK: call void @__kmpc_critical( 222 // CHECK: call {{.*}}void @_Z3redR6BaseS1RKS_( 223 // CHECK: call void @__kmpc_end_critical( 224 225 // var1 = var1.operator &&(var1_reduction); 226 // CHECK: call void @__kmpc_critical( 227 // CHECK: fmul float 228 // CHECK: call void @__kmpc_end_critical( 229 230 // t_var1 = min(t_var1, t_var1_reduction); 231 // CHECK: call void @__kmpc_critical( 232 // CHECK: fadd float 5.550000e+02, % 233 // CHECK: call void @__kmpc_end_critical( 234 235 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 236 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 237 238 // break; 239 // CHECK: br label %[[RED_DONE]] 240 // CHECK: [[RED_DONE]] 241 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 242 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* 243 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 244 245 // CHECK: ret void 246 247 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 248 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 249 // ... 250 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 251 // *(Type<n>-1*)rhs[<n>-1]); 252 // } 253 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 254 // t_var_lhs = (float*)lhs[0]; 255 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 256 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 257 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to float* 258 // t_var_rhs = (float*)rhs[0]; 259 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 260 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 261 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to float* 262 263 // var_lhs = (S<float>*)lhs[1]; 264 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 265 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 266 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_FLOAT_TY]]* 267 // var_rhs = (S<float>*)rhs[1]; 268 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 269 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 270 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_FLOAT_TY]]* 271 272 // var1_lhs = (S<float>*)lhs[2]; 273 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 274 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 275 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_FLOAT_TY]]* 276 // var1_rhs = (S<float>*)rhs[2]; 277 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 278 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 279 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_FLOAT_TY]]* 280 281 // t_var1_lhs = (float*)lhs[3]; 282 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 283 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 284 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to float* 285 // t_var1_rhs = (float*)rhs[3]; 286 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 287 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 288 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to float* 289 290 // t_var_lhs += t_var_rhs; 291 // CHECK: fsub float 2.220000e+02, % 292 293 // var_lhs = var_lhs.operator &(var_rhs); 294 // CHECK: call {{.*}}void @_Z3redR6BaseS1RKS_( 295 296 // var1_lhs = var1_lhs.operator &&(var1_rhs); 297 // CHECK: fmul float 298 299 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 300 // CHECK: fadd float 5.550000e+02, % 301 // CHECK: ret void 302 303 // CHECK: define internal void [[MAIN_MICROTASK1]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i64 %{{.+}}, i64 %{{.+}}, i32* %{{.+}}, [2 x i32]* dereferenceable(8) %{{.+}}, [10 x [4 x [[S_FLOAT_TY]]]]* dereferenceable(480) %{{.+}}) 304 305 // Reduction list for runtime. 306 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 307 308 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 309 // CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}** 310 // CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}** [[ARR_ADDR:%.+]], 311 // CHECK: [[ARR:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[ARR_ADDR]], 312 313 // CHECK: [[UB_CAST:%.+]] = ptrtoint i32* [[UB1_UP:%.+]] to i64 314 // CHECK: [[LB_CAST:%.+]] = ptrtoint i32* [[LB1_0:%.+]] to i64 315 // CHECK: [[DIFF:%.+]] = sub i64 [[UB_CAST]], [[LB_CAST]] 316 // CHECK: [[SIZE_1:%.+]] = sdiv exact i64 [[DIFF]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64) 317 // CHECK: [[ARR_SIZE:%.+]] = add nuw i64 [[SIZE_1]], 1 318 // CHECK: call i8* @llvm.stacksave() 319 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]], 320 321 // Check initialization of private copy. 322 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]] 323 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]] 324 // CHECK: br i1 [[ISEMPTY]], 325 // CHECK: phi i32* 326 // CHECK: store i32 888, i32* % 327 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 328 // CHECK: br i1 [[DONE]], 329 330 // CHECK: [[ARRS_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[ARRS_SIZE:%.+]], 331 332 // Check initialization of private copy. 333 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]] 334 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]] 335 // CHECK: br i1 [[ISEMPTY]], 336 // CHECK: phi [[S_FLOAT_TY]]* 337 // CHECK: call void @_Z4initR6BaseS1RKS_(% 338 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 339 // CHECK: br i1 [[DONE]], 340 341 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 342 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 343 // CHECK: call void @__kmpc_for_static_init_4( 344 // Skip checks for internal operations. 345 // CHECK: call void @__kmpc_for_static_fini( 346 347 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 348 349 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 350 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8* 351 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]], 352 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 353 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8* 354 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]], 355 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 356 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV]] to i8* 357 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]], 358 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 359 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARRS_SIZE]] to i8* 360 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_SIZE_REF]], 361 362 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 363 364 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 365 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 366 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 367 // 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]]) 368 369 // switch(res) 370 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 371 // CHECK: i32 1, label %[[CASE1:.+]] 372 // CHECK: i32 2, label %[[CASE2:.+]] 373 // CHECK: ] 374 375 // case 1: 376 // CHECK: [[CASE1]] 377 378 // arr[:] += arr_reduction[:]; 379 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 380 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 381 // CHECK: br i1 [[ISEMPTY]], 382 // CHECK: phi i32* 383 // CHECK: [[ADD:%.+]] = mul nsw i32 555, % 384 // CHECK: store i32 [[ADD]], i32* % 385 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 386 // CHECK: br i1 [[DONE]], 387 388 // arrs[:] = var.operator &(arrs_reduction[:]); 389 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 390 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 391 // CHECK: br i1 [[ISEMPTY]], 392 // CHECK: phi [[S_FLOAT_TY]]* 393 // CHECK: call void @_Z3redR6BaseS1RKS_(% 394 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 395 // CHECK: br i1 [[DONE]], 396 397 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 398 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 399 400 // break; 401 // CHECK: br label %[[RED_DONE]] 402 403 // case 2: 404 // CHECK: [[CASE2]] 405 406 // arr[:] += arr_reduction[:]; 407 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 408 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 409 // CHECK: br i1 [[ISEMPTY]], 410 // CHECK: phi i32* 411 // CHECK: call void @__kmpc_critical( 412 // CHECK: [[ADD:%.+]] = mul nsw i32 555, % 413 // CHECK: call void @__kmpc_end_critical( 414 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 415 // CHECK: br i1 [[DONE]], 416 417 // arrs[:] = var.operator &(arrs_reduction[:]); 418 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 419 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 420 // CHECK: br i1 [[ISEMPTY]], 421 // CHECK: phi [[S_FLOAT_TY]]* 422 // CHECK: call void @__kmpc_critical( 423 // CHECK: call void @_Z3redR6BaseS1RKS_(% 424 // CHECK: call void @__kmpc_end_critical( 425 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 426 // CHECK: br i1 [[DONE]], 427 428 // break; 429 // CHECK: br label %[[RED_DONE]] 430 // CHECK: [[RED_DONE]] 431 432 // Check destruction of private copy. 433 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]] 434 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]] 435 // CHECK: br i1 [[ISEMPTY]], 436 // CHECK: phi [[S_FLOAT_TY]]* 437 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* % 438 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[ARRS_PRIV]] 439 // CHECK: br i1 [[DONE]], 440 // CHECK: call void @llvm.stackrestore(i8* 441 442 // CHECK: ret void 443 444 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 445 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 446 // ... 447 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 448 // *(Type<n>-1*)rhs[<n>-1]); 449 // } 450 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 451 // arr_rhs = (int*)rhs[0]; 452 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 453 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]], 454 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32* 455 // arr_lhs = (int*)lhs[0]; 456 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 457 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]], 458 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32* 459 460 // arr_size = (size_t)lhs[1]; 461 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 462 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]], 463 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64 464 465 // arrs_rhs = (S<float>*)rhs[2]; 466 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 467 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]], 468 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]* 469 // arrs_lhs = (S<float>*)lhs[2]; 470 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 471 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]], 472 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]* 473 474 // arrs_size = (size_t)lhs[3]; 475 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 476 // CHECK: [[ARRS_SIZE_VOID:%.+]] = load i8*, i8** [[ARRS_SIZE_REF]], 477 // CHECK: [[ARRS_SIZE:%.+]] = ptrtoint i8* [[ARRS_SIZE_VOID]] to i64 478 479 // arr_lhs[:] += arr_rhs[:]; 480 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]] 481 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]] 482 // CHECK: br i1 [[ISEMPTY]], 483 // CHECK: phi i32* 484 // CHECK: [[ADD:%.+]] = mul nsw i32 555, % 485 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 486 // CHECK: br i1 [[DONE]], 487 488 // arrs_lhs = arrs_lhs.operator &(arrs_rhs); 489 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 490 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 491 // CHECK: br i1 [[ISEMPTY]], 492 // CHECK: phi [[S_FLOAT_TY]]* 493 // CHECK: call void @_Z3redR6BaseS1RKS_(% 494 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 495 // CHECK: br i1 [[DONE]], 496 497 // CHECK: ret void 498 499 // CHECK: define internal void [[MAIN_MICROTASK2]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i64 %{{.+}}, i64 %{{.+}}, i32* %{{.+}}, [10 x [4 x [[S_FLOAT_TY]]]]* dereferenceable(480) %{{.+}}) 500 501 // CHECK: [[ARRS_PRIV:%.+]] = alloca [10 x [4 x [[S_FLOAT_TY]]]], 502 503 // Reduction list for runtime. 504 // CHECK: [[RED_LIST:%.+]] = alloca [3 x i8*], 505 506 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 507 508 // CHECK: [[ARR_SIZE:%.+]] = udiv exact i64 509 // CHECK: call i8* @llvm.stacksave() 510 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]], 511 512 // Check initialization of private copy. 513 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]] 514 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]] 515 // CHECK: br i1 [[ISEMPTY]], 516 // CHECK: phi i32* 517 // CHECK: store i32 888, i32* % 518 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 519 // CHECK: br i1 [[DONE]], 520 521 // Check initialization of private copy. 522 // CHECK: [[BEGIN:%.+]] = getelementptr inbounds [10 x [4 x [[S_FLOAT_TY]]]], [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]], i32 0, i32 0, i32 0 523 // CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]* 524 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40 525 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[BEGIN]], [[END]] 526 // CHECK: br i1 [[ISEMPTY]], 527 // CHECK: phi [[S_FLOAT_TY]]* 528 // CHECK: call void @_Z4initR6BaseS1RKS_(% 529 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 530 // CHECK: br i1 [[DONE]], 531 // CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]* 532 // CHECK: [[ARRS_PRIV_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]] to [[S_FLOAT_TY]]* 533 534 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 535 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 536 // CHECK: call void @__kmpc_for_static_init_4( 537 // Skip checks for internal operations. 538 // CHECK: call void @__kmpc_for_static_fini( 539 540 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 541 542 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 0 543 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8* 544 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]], 545 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 1 546 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8* 547 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]], 548 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 2 549 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV_BEGIN]] to i8* 550 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]], 551 552 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 553 554 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 555 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 556 // CHECK: [[BITCAST:%.+]] = bitcast [3 x i8*]* [[RED_LIST]] to i8* 557 // CHECK: [[RES:%.+]] = call i32 @__kmpc_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], i32 2, i64 24, i8* [[BITCAST]], void (i8*, i8*)* [[REDUCTION_FUNC:@.+]], [8 x i32]* [[REDUCTION_LOCK]]) 558 559 // switch(res) 560 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 561 // CHECK: i32 1, label %[[CASE1:.+]] 562 // CHECK: i32 2, label %[[CASE2:.+]] 563 // CHECK: ] 564 565 // case 1: 566 // CHECK: [[CASE1]] 567 568 // arr[:] += arr_reduction[:]; 569 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0:%.+]], i64 [[ARR_SIZE]] 570 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 571 // CHECK: br i1 [[ISEMPTY]], 572 // CHECK: phi i32* 573 // CHECK: [[ADD:%[^ ]+]] = mul nsw i32 555, % 574 // CHECK: store i32 [[ADD]], i32* % 575 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 576 // CHECK: br i1 [[DONE]], 577 578 // arrs[:] = var.operator &(arrs_reduction[:]); 579 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40 580 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]] 581 // CHECK: br i1 [[ISEMPTY]], 582 // CHECK: phi [[S_FLOAT_TY]]* 583 // CHECK: call void @_Z3redR6BaseS1RKS_(% 584 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 585 // CHECK: br i1 [[DONE]], 586 587 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 588 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 589 590 // break; 591 // CHECK: br label %[[RED_DONE]] 592 593 // case 2: 594 // CHECK: [[CASE2]] 595 596 // arr[:] += arr_reduction[:]; 597 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 598 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 599 // CHECK: br i1 [[ISEMPTY]], 600 // CHECK: phi i32* 601 // CHECK: call void @__kmpc_critical( 602 // CHECK: [[ADD:%.+]] = mul nsw i32 555, % 603 // CHECK: call void @__kmpc_end_critical( 604 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 605 // CHECK: br i1 [[DONE]], 606 607 // arrs[:] = var.operator &(arrs_reduction[:]); 608 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40 609 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]] 610 // CHECK: br i1 [[ISEMPTY]], 611 // CHECK: phi [[S_FLOAT_TY]]* 612 // CHECK: call void @__kmpc_critical( 613 // CHECK: call void @_Z3redR6BaseS1RKS_(% 614 // CHECK: call void @__kmpc_end_critical( 615 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 616 // CHECK: br i1 [[DONE]], 617 618 // break; 619 // CHECK: br label %[[RED_DONE]] 620 // CHECK: [[RED_DONE]] 621 622 // Check destruction of private copy. 623 // CHECK: [[BEGIN:%.+]] = getelementptr inbounds [10 x [4 x [[S_FLOAT_TY]]]], [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]], i32 0, i32 0, i32 0 624 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40 625 // CHECK: br 626 // CHECK: phi [[S_FLOAT_TY]]* 627 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* % 628 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[BEGIN]] 629 // CHECK: br i1 [[DONE]], 630 // CHECK: call void @llvm.stackrestore(i8* 631 // CHECK: call void @__kmpc_barrier( 632 633 // CHECK: ret void 634 635 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 636 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 637 // ... 638 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 639 // *(Type<n>-1*)rhs[<n>-1]); 640 // } 641 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 642 // arr_rhs = (int*)rhs[0]; 643 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 644 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]], 645 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32* 646 // arr_lhs = (int*)lhs[0]; 647 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 648 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]], 649 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32* 650 651 // arr_size = (size_t)lhs[1]; 652 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 653 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]], 654 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64 655 656 // arrs_rhs = (S<float>*)rhs[2]; 657 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 658 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]], 659 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]* 660 // arrs_lhs = (S<float>*)lhs[2]; 661 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 662 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]], 663 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]* 664 665 // arr_lhs[:] += arr_rhs[:]; 666 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]] 667 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]] 668 // CHECK: br i1 [[ISEMPTY]], 669 // CHECK: phi i32* 670 // CHECK: [[ADD:%.+]] = mul nsw i32 555, % 671 // CHECK: store i32 [[ADD]], i32* % 672 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 673 // CHECK: br i1 [[DONE]], 674 675 // arrs_lhs = arrs_lhs.operator &(arrs_rhs); 676 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 40 677 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 678 // CHECK: br i1 [[ISEMPTY]], 679 // CHECK: phi [[S_FLOAT_TY]]* 680 // CHECK: call void @_Z3redR6BaseS1RKS_(% 681 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 682 // CHECK: br i1 [[DONE]], 683 684 // CHECK: ret void 685 686 // CHECK: define internal void [[MAIN_MICROTASK3]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** dereferenceable(8) %{{.+}}) 687 688 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***, 689 690 // Reduction list for runtime. 691 // CHECK: [[RED_LIST:%.+]] = alloca [2 x i8*], 692 693 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 694 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]], 695 696 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64 697 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW:%.+]] to i64 698 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]] 699 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64) 700 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1 701 // CHECK: call i8* @llvm.stacksave() 702 // CHECK: [[VAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]], 703 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 704 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], 705 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 706 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 707 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 708 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64) 709 // CHECK: [[PSEUDO_VAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR2_PRIV]], i64 [[OFFSET]] 710 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** % 711 // CHECK: store [[S_FLOAT_TY]]* [[PSEUDO_VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]] 712 // CHECK: ret void 713 714 // CHECK: define internal void [[MAIN_MICROTASK4]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [2 x [[S_FLOAT_TY]]]* dereferenceable(24) %{{.+}}) 715 716 // CHECK: [[VVAR2_ORIG_ADDR:%.+]] = alloca [2 x [[S_FLOAT_TY]]]*, 717 718 // Reduction list for runtime. 719 // CHECK: [[RED_LIST:%.+]] = alloca [2 x i8*], 720 721 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 722 // CHECK: [[VVAR2_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VVAR2_ORIG_ADDR]], 723 724 // CHECK: [[LOW:%.+]] = getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]], i64 0, i64 0 725 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64 726 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 727 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]] 728 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64) 729 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1 730 // CHECK: call i8* @llvm.stacksave() 731 // CHECK: [[VVAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]], 732 // CHECK: [[ORIG_START:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]] to [[S_FLOAT_TY]]* 733 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 734 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 735 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 736 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64) 737 // CHECK: [[PSEUDO_VVAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VVAR2_PRIV]], i64 [[OFFSET]] 738 // CHECK: [[VVAR2_PRIV:%.+]] = bitcast [[S_FLOAT_TY]]* [[PSEUDO_VVAR2_PRIV]] to [2 x [[S_FLOAT_TY]]]* 739 // CHECK: ret void 740 741 // CHECK: define internal void [[MAIN_MICROTASK5]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [2 x [[S_FLOAT_TY]]]* dereferenceable(24) %{{.+}}) 742 743 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [2 x [[S_FLOAT_TY]]]*, 744 745 // Reduction list for runtime. 746 // CHECK: [[RED_LIST:%.+]] = alloca [2 x i8*], 747 748 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 749 750 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 751 // CHECK: store [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 752 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64 753 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW:%.+]] to i64 754 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]] 755 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64) 756 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1 757 // CHECK: call i8* @llvm.stacksave() 758 // CHECK: [[VAR3_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]], 759 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 760 // CHECK: [[ORIG_START:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 761 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 762 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 763 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 764 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64) 765 // CHECK: [[PSEUDO_VAR3_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR3_PRIV]], i64 [[OFFSET]] 766 // CHECK: [[VAR3_PRIV:%.+]] = bitcast [[S_FLOAT_TY]]* [[PSEUDO_VAR3_PRIV]] to [2 x [[S_FLOAT_TY]]]* 767 768 // CHECK: store [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [2 x [[S_FLOAT_TY]]]** % 769 770 // CHECK: ret void 771 772 // CHECK: define internal void [[MAIN_MICROTASK6]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [2 x [[S_FLOAT_TY]]]* dereferenceable(24) %{{.+}}) 773 774 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [2 x [[S_FLOAT_TY]]]*, 775 // CHECK: [[VAR3_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]], 776 777 // Reduction list for runtime. 778 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 779 780 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 781 782 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 783 // CHECK: store [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 784 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 785 // CHECK: getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], i32 0, i32 0 786 // CHECK: bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 787 // CHECK: getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 2 788 789 // CHECK: store [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [2 x [[S_FLOAT_TY]]]** % 790 791 // CHECK: ret void 792 793 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT]]() 794 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]], 795 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]]) 796 // 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 797 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]* 798 // CHECK: ret 799 // 800 // CHECK: define internal void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i32* dereferenceable(4) %{{.+}}, [[S_INT_TY]]* dereferenceable(12) %{{.+}}, [[S_INT_TY]]* dereferenceable(12) %{{.+}}, i32* dereferenceable(4) %{{.+}}, [2 x i32]* dereferenceable(8) %{{.+}}, [2 x [[S_INT_TY]]]* dereferenceable(24) %{{.+}}) 801 // CHECK: alloca i{{[0-9]+}}, 802 // CHECK: alloca i{{[0-9]+}}, 803 // CHECK: alloca i{{[0-9]+}}, 804 // CHECK: alloca i{{[0-9]+}}, 805 // CHECK: alloca i{{[0-9]+}}, 806 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 807 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], 808 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_INT_TY]], 809 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca i{{[0-9]+}}, 810 811 // Reduction list for runtime. 812 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 813 814 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 815 816 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 817 // CHECK: [[VAR1_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 818 // CHECK: [[T_VAR1_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 819 820 // For + reduction operation initial value of private variable is 0. 821 // CHECK: store i32 321, i32* % 822 823 // For & reduction operation initial value of private variable is ones in all bits. 824 // CHECK: call void @_Z4initR6BaseS1RKS_( 825 826 // For && reduction operation initial value of private variable is 1.0. 827 // CHECK: call void @_Z5init2R6BaseS1RKS_( 828 829 // For min reduction operation initial value of private variable is largest repesentable value. 830 // CHECK: sdiv i32 432, % 831 832 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 833 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 834 // CHECK: call void @__kmpc_for_static_init_4( 835 // Skip checks for internal operations. 836 // CHECK: call void @__kmpc_for_static_fini( 837 838 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 839 840 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 841 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR_PRIV]] to i8* 842 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 843 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 844 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR_PRIV]] to i8* 845 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 846 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 847 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_PRIV]] to i8* 848 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 849 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 850 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR1_PRIV]] to i8* 851 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 852 853 // res = __kmpc_reduce_nowait(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 854 855 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 856 // 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]]) 857 858 // switch(res) 859 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 860 // CHECK: i32 1, label %[[CASE1:.+]] 861 // CHECK: i32 2, label %[[CASE2:.+]] 862 // CHECK: ] 863 864 // case 1: 865 // t_var += t_var_reduction; 866 // CHECK: add nsw i32 1513, % 867 868 // var = var.operator &(var_reduction); 869 // CHECK: call void @_Z3redR6BaseS1RKS_(% 870 871 // var1 = var1.operator &&(var1_reduction); 872 // CHECK: mul nsw i32 17, % 873 874 // t_var1 = min(t_var1, t_var1_reduction); 875 // CHECK: sub nsw i32 47, % 876 877 // __kmpc_end_reduce_nowait(<loc>, <gtid>, &<lock>); 878 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 879 880 // break; 881 // CHECK: br label %[[RED_DONE]] 882 883 // case 2: 884 // t_var += t_var_reduction; 885 // CHECK: call void @__kmpc_critical( 886 // CHECK: add nsw i32 1513, % 887 // CHECK: call void @__kmpc_end_critical( 888 889 // var = var.operator &(var_reduction); 890 // CHECK: call void @__kmpc_critical( 891 // CHECK: call void @_Z3redR6BaseS1RKS_(% 892 // CHECK: call void @__kmpc_end_critical( 893 894 // var1 = var1.operator &&(var1_reduction); 895 // CHECK: call void @__kmpc_critical( 896 // CHECK: mul nsw i32 17, % 897 // CHECK: call void @__kmpc_end_critical( 898 899 // t_var1 = min(t_var1, t_var1_reduction); 900 // CHECK: call void @__kmpc_critical( 901 // CHECK: sub nsw i32 47, % 902 // CHECK: call void @__kmpc_end_critical( 903 904 // break; 905 // CHECK: br label %[[RED_DONE]] 906 // CHECK: [[RED_DONE]] 907 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) 908 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* 909 // CHECK: ret void 910 911 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 912 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 913 // ... 914 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 915 // *(Type<n>-1*)rhs[<n>-1]); 916 // } 917 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 918 // t_var_lhs = (i{{[0-9]+}}*)lhs[0]; 919 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 920 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 921 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to i{{[0-9]+}}* 922 // t_var_rhs = (i{{[0-9]+}}*)rhs[0]; 923 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 924 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 925 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to i{{[0-9]+}}* 926 927 // var_lhs = (S<i{{[0-9]+}}>*)lhs[1]; 928 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 929 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 930 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_INT_TY]]* 931 // var_rhs = (S<i{{[0-9]+}}>*)rhs[1]; 932 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 933 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 934 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_INT_TY]]* 935 936 // var1_lhs = (S<i{{[0-9]+}}>*)lhs[2]; 937 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 938 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 939 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_INT_TY]]* 940 // var1_rhs = (S<i{{[0-9]+}}>*)rhs[2]; 941 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 942 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 943 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_INT_TY]]* 944 945 // t_var1_lhs = (i{{[0-9]+}}*)lhs[3]; 946 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 947 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 948 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to i{{[0-9]+}}* 949 // t_var1_rhs = (i{{[0-9]+}}*)rhs[3]; 950 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 951 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 952 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to i{{[0-9]+}}* 953 954 // t_var_lhs += t_var_rhs; 955 // CHECK: add nsw i32 1513, % 956 957 // var_lhs = var_lhs.operator &(var_rhs); 958 // CHECK: call void @_Z3redR6BaseS1RKS_(% 959 960 // var1_lhs = var1_lhs.operator &&(var1_rhs); 961 // CHECK: mul nsw i32 17, % 962 963 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 964 // CHECK: sub nsw i32 47, % 965 // CHECK: ret void 966 967 #endif 968 969