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 5 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s 6 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s 7 // RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY0 %s 8 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}} 9 // expected-no-diagnostics 10 #ifndef HEADER 11 #define HEADER 12 13 volatile double g, g_orig; 14 volatile double &g1 = g_orig; 15 16 struct BaseS { 17 int x; 18 }; 19 struct BaseS1 { 20 float y; 21 }; 22 23 template <class T> 24 struct S : public BaseS, public BaseS1 { 25 T f; 26 S(T a) : f(a + g) {} 27 S() : f(g) {} 28 ~S() {} 29 }; 30 void red(BaseS1&, const BaseS1&); 31 void red_plus(BaseS1&, const BaseS1&); 32 void init(BaseS1&, const BaseS1&); 33 void init1(BaseS1&, const BaseS1&); 34 void init2(BaseS1&, const BaseS1&); 35 void init_plus(BaseS1&, const BaseS1&); 36 #pragma omp declare reduction(operator& : BaseS1 : red(omp_out, omp_in)) initializer(init(omp_priv, omp_orig)) 37 #pragma omp declare reduction(+ : BaseS1 : red_plus(omp_out, omp_in)) initializer(init_plus(omp_priv, omp_orig)) 38 #pragma omp declare reduction(&& : S<float>, S<int> : omp_out.f *= omp_in.f) initializer(init1(omp_priv, omp_orig)) 39 40 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { %{{[^,]+}}, %{{[^,]+}}, float } 41 // CHECK-DAG: [[S_INT_TY:%.+]] = type { %{{[^,]+}}, %{{[^,]+}}, i{{[0-9]+}} } 42 // CHECK-DAG: [[ATOMIC_REDUCE_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 43 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8* 44 // CHECK-DAG: [[REDUCTION_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 45 // CHECK-DAG: [[REDUCTION_LOCK:@.+]] = common global [8 x i32] zeroinitializer 46 47 #pragma omp declare reduction(operator&& : int : omp_out = 111 & omp_in) 48 template <typename T, int length> 49 T tmain() { 50 T t; 51 S<T> test; 52 T t_var = T(), t_var1; 53 T vec[] = {1, 2}; 54 S<T> s_arr[] = {1, 2}; 55 S<T> &var = test; 56 S<T> var1; 57 S<T> arr[length]; 58 #pragma omp declare reduction(operator& : T : omp_out = 15 + omp_in) 59 #pragma omp declare reduction(operator+ : T : omp_out = 1513 + omp_in) initializer(omp_priv = 321) 60 #pragma omp declare reduction(min : T : omp_out = 47 - omp_in) initializer(omp_priv = 432 / omp_orig) 61 #pragma omp declare reduction(operator&& : S<T> : omp_out.f = 17 * omp_in.f) initializer(init2(omp_priv, omp_orig)) 62 #pragma omp declare reduction(operator&& : T : omp_out = 17 * omp_in) 63 #pragma omp parallel 64 #pragma omp for reduction(+ : t_var) reduction(& : var) reduction(&& : var1) reduction(min : t_var1) nowait 65 for (int i = 0; i < 2; ++i) { 66 vec[i] = t_var; 67 s_arr[i] = var; 68 } 69 #pragma omp parallel 70 #pragma omp for reduction(&& : t_var) 71 for (int i = 0; i < 2; ++i) { 72 vec[i] = t_var; 73 s_arr[i] = var; 74 } 75 #pragma omp parallel 76 #pragma omp for reduction(+ : arr[1:length-2]) 77 for (int i = 0; i < 2; ++i) { 78 vec[i] = t_var; 79 s_arr[i] = var; 80 } 81 return T(); 82 } 83 84 extern S<float> **foo(); 85 86 #pragma omp declare reduction(operator- : float, double : omp_out = 333 + omp_in) 87 #pragma omp declare reduction(min : float, double : omp_out = 555 + omp_in) 88 int main() { 89 #pragma omp declare reduction(operator+ : float, double : omp_out = 222 - omp_in) initializer(omp_priv = -1) 90 S<float> test; 91 float t_var = 0, t_var1; 92 int vec[] = {1, 2}; 93 S<float> s_arr[] = {1, 2, 3, 4}; 94 S<float> &var = test; 95 S<float> var1, arrs[10][4]; 96 S<float> **var2 = foo(); 97 S<float> vvar2[5]; 98 S<float>(&var3)[4] = s_arr; 99 #pragma omp declare reduction(operator+ : int : omp_out = 555 * omp_in) initializer(omp_priv = 888) 100 #pragma omp parallel 101 #pragma omp for reduction(+ : t_var) reduction(& : var) reduction(&& : var1) reduction(min : t_var1) 102 for (int i = 0; i < 2; ++i) { 103 vec[i] = t_var; 104 s_arr[i] = var; 105 } 106 int arr[10][vec[1]]; 107 #pragma omp parallel for reduction(+ : arr[1][ : vec[1]]) reduction(& : arrs[1 : vec[1]][1 : 2]) 108 for (int i = 0; i < 10; ++i) 109 ++arr[1][i]; 110 #pragma omp parallel 111 #pragma omp for reduction(+ : arr) reduction(& : arrs) 112 for (int i = 0; i < 10; ++i) 113 ++arr[1][i]; 114 #pragma omp parallel 115 #pragma omp for reduction(& : var2[0 : 5][1 : 6]) 116 for (int i = 0; i < 10; ++i) 117 ; 118 #pragma omp parallel 119 #pragma omp for reduction(& : vvar2[0 : 5]) 120 for (int i = 0; i < 10; ++i) 121 ; 122 #pragma omp parallel 123 #pragma omp for reduction(& : var3[1 : 2]) 124 for (int i = 0; i < 10; ++i) 125 ; 126 #pragma omp parallel 127 #pragma omp for reduction(& : var3) 128 for (int i = 0; i < 10; ++i) 129 ; 130 return tmain<int, 42>(); 131 } 132 133 // CHECK: define {{.*}}i{{[0-9]+}} @main() 134 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]], 135 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]]) 136 // 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]*, [4 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK:@.+]] to void 137 // 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 138 // 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 139 // 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 140 // 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]+}}*, [5 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK4:@.+]] to void 141 // 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]+}}*, [4 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK5:@.+]] to void 142 // 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]+}}*, [4 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK6:@.+]] to void 143 // CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT_42:@.+]]() 144 // CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]* 145 // CHECK: ret 146 // 147 // 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, [4 x [[S_FLOAT_TY]]]* dereferenceable(48) %{{.+}}) 148 // CHECK: [[T_VAR_PRIV:%.+]] = alloca float, 149 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 150 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 151 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca float, 152 153 // Reduction list for runtime. 154 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 155 156 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 157 158 // CHECK: [[T_VAR_REF:%.+]] = load float*, float** % 159 // CHECK: [[VAR1_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** % 160 // CHECK: [[T_VAR1_REF:%.+]] = load float*, float** % 161 162 // For + reduction operation initial value of private variable is -1. 163 // CHECK: store float -1.0{{.+}}, float* 164 165 // For & reduction operation initial value of private variable is defined by call of 'init()' function. 166 // CHECK: call {{.*}}void @_Z4initR6BaseS1RKS_( 167 168 // For && reduction operation initial value of private variable is 1.0. 169 // CHECK: call {{.*}}void @_Z5init1R6BaseS1RKS_( 170 171 // For min reduction operation initial value of private variable is largest repesentable value. 172 // CHECK: [[INIT:%.+]] = load float, float* @ 173 // CHECK: store float [[INIT]], float* [[T_VAR1_PRIV]], 174 175 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 176 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 177 // CHECK: call void @__kmpc_for_static_init_4( 178 // Skip checks for internal operations. 179 // CHECK: call void @__kmpc_for_static_fini( 180 181 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 182 183 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 184 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR_PRIV]] to i8* 185 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 186 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 187 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_PRIV]] to i8* 188 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 189 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 190 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_PRIV]] to i8* 191 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 192 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 193 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR1_PRIV]] to i8* 194 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 195 196 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 197 198 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 199 // 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]]) 200 201 // switch(res) 202 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 203 // CHECK: i32 1, label %[[CASE1:.+]] 204 // CHECK: i32 2, label %[[CASE2:.+]] 205 // CHECK: ] 206 207 // case 1: 208 // t_var += t_var_reduction; 209 // CHECK: fsub float 2.220000e+02, % 210 211 // var = var.operator &(var_reduction); 212 // CHECK: call {{.*}}void @_Z3redR6BaseS1RKS_( 213 214 // var1 = var1.operator &&(var1_reduction); 215 // CHECK: fmul float 216 217 // t_var1 = min(t_var1, t_var1_reduction); 218 // CHECK: fadd float 5.550000e+02, % 219 220 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 221 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 222 223 // break; 224 // CHECK: br label %[[RED_DONE]] 225 226 // case 2: 227 // t_var += t_var_reduction; 228 // CHECK: call void @__kmpc_critical( 229 // CHECK: fsub float 2.220000e+02, % 230 // CHECK: call void @__kmpc_end_critical( 231 232 // var = var.operator &(var_reduction); 233 // CHECK: call void @__kmpc_critical( 234 // CHECK: call {{.*}}void @_Z3redR6BaseS1RKS_( 235 // CHECK: call void @__kmpc_end_critical( 236 237 // var1 = var1.operator &&(var1_reduction); 238 // CHECK: call void @__kmpc_critical( 239 // CHECK: fmul float 240 // CHECK: call void @__kmpc_end_critical( 241 242 // t_var1 = min(t_var1, t_var1_reduction); 243 // CHECK: call void @__kmpc_critical( 244 // CHECK: fadd float 5.550000e+02, % 245 // CHECK: call void @__kmpc_end_critical( 246 247 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 248 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 249 250 // break; 251 // CHECK: br label %[[RED_DONE]] 252 // CHECK: [[RED_DONE]] 253 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 254 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* 255 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 256 257 // CHECK: ret void 258 259 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 260 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 261 // ... 262 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 263 // *(Type<n>-1*)rhs[<n>-1]); 264 // } 265 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 266 // t_var_lhs = (float*)lhs[0]; 267 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 268 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 269 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to float* 270 // t_var_rhs = (float*)rhs[0]; 271 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 272 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 273 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to float* 274 275 // var_lhs = (S<float>*)lhs[1]; 276 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 277 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 278 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_FLOAT_TY]]* 279 // var_rhs = (S<float>*)rhs[1]; 280 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 281 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 282 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_FLOAT_TY]]* 283 284 // var1_lhs = (S<float>*)lhs[2]; 285 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 286 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 287 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_FLOAT_TY]]* 288 // var1_rhs = (S<float>*)rhs[2]; 289 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 290 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 291 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_FLOAT_TY]]* 292 293 // t_var1_lhs = (float*)lhs[3]; 294 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 295 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 296 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to float* 297 // t_var1_rhs = (float*)rhs[3]; 298 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 299 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 300 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to float* 301 302 // t_var_lhs += t_var_rhs; 303 // CHECK: fsub float 2.220000e+02, % 304 305 // var_lhs = var_lhs.operator &(var_rhs); 306 // CHECK: call {{.*}}void @_Z3redR6BaseS1RKS_( 307 308 // var1_lhs = var1_lhs.operator &&(var1_rhs); 309 // CHECK: fmul float 310 311 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 312 // CHECK: fadd float 5.550000e+02, % 313 // CHECK: ret void 314 315 // 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) %{{.+}}) 316 317 // Reduction list for runtime. 318 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 319 320 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 321 // CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}** 322 // CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}** [[ARR_ADDR:%.+]], 323 // CHECK: [[ARR:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[ARR_ADDR]], 324 325 // CHECK: [[UB_CAST:%.+]] = ptrtoint i32* [[UB1_UP:%.+]] to i64 326 // CHECK: [[LB_CAST:%.+]] = ptrtoint i32* [[LB1_0:%.+]] to i64 327 // CHECK: [[DIFF:%.+]] = sub i64 [[UB_CAST]], [[LB_CAST]] 328 // CHECK: [[SIZE_1:%.+]] = sdiv exact i64 [[DIFF]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64) 329 // CHECK: [[ARR_SIZE:%.+]] = add nuw i64 [[SIZE_1]], 1 330 // CHECK: call i8* @llvm.stacksave() 331 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]], 332 333 // Check initialization of private copy. 334 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]] 335 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]] 336 // CHECK: br i1 [[ISEMPTY]], 337 // CHECK: phi i32* 338 // CHECK: store i32 888, i32* % 339 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 340 // CHECK: br i1 [[DONE]], 341 342 // CHECK: [[ARRS_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[ARRS_SIZE:%.+]], 343 344 // Check initialization of private copy. 345 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]] 346 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]] 347 // CHECK: br i1 [[ISEMPTY]], 348 // CHECK: phi [[S_FLOAT_TY]]* 349 // CHECK: call void @_Z4initR6BaseS1RKS_(% 350 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 351 // CHECK: br i1 [[DONE]], 352 353 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 354 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 355 // CHECK: call void @__kmpc_for_static_init_4( 356 // Skip checks for internal operations. 357 // CHECK: call void @__kmpc_for_static_fini( 358 359 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 360 361 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 362 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8* 363 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]], 364 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 365 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8* 366 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]], 367 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 368 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV]] to i8* 369 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]], 370 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 371 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARRS_SIZE]] to i8* 372 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_SIZE_REF]], 373 374 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 375 376 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 377 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 378 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 379 // 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]]) 380 381 // switch(res) 382 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 383 // CHECK: i32 1, label %[[CASE1:.+]] 384 // CHECK: i32 2, label %[[CASE2:.+]] 385 // CHECK: ] 386 387 // case 1: 388 // CHECK: [[CASE1]] 389 390 // arr[:] += arr_reduction[:]; 391 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 392 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 393 // CHECK: br i1 [[ISEMPTY]], 394 // CHECK: phi i32* 395 // CHECK: [[ADD:%.+]] = mul nsw i32 555, % 396 // CHECK: store i32 [[ADD]], i32* % 397 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 398 // CHECK: br i1 [[DONE]], 399 400 // arrs[:] = var.operator &(arrs_reduction[:]); 401 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 402 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 403 // CHECK: br i1 [[ISEMPTY]], 404 // CHECK: phi [[S_FLOAT_TY]]* 405 // CHECK: call void @_Z3redR6BaseS1RKS_(% 406 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 407 // CHECK: br i1 [[DONE]], 408 409 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 410 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 411 412 // break; 413 // CHECK: br label %[[RED_DONE]] 414 415 // case 2: 416 // CHECK: [[CASE2]] 417 418 // arr[:] += arr_reduction[:]; 419 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 420 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 421 // CHECK: br i1 [[ISEMPTY]], 422 // CHECK: phi i32* 423 // CHECK: call void @__kmpc_critical( 424 // CHECK: [[ADD:%.+]] = mul nsw i32 555, % 425 // CHECK: call void @__kmpc_end_critical( 426 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 427 // CHECK: br i1 [[DONE]], 428 429 // arrs[:] = var.operator &(arrs_reduction[:]); 430 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 431 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 432 // CHECK: br i1 [[ISEMPTY]], 433 // CHECK: phi [[S_FLOAT_TY]]* 434 // CHECK: call void @__kmpc_critical( 435 // CHECK: call void @_Z3redR6BaseS1RKS_(% 436 // CHECK: call void @__kmpc_end_critical( 437 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 438 // CHECK: br i1 [[DONE]], 439 440 // break; 441 // CHECK: br label %[[RED_DONE]] 442 // CHECK: [[RED_DONE]] 443 444 // Check destruction of private copy. 445 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]] 446 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]] 447 // CHECK: br i1 [[ISEMPTY]], 448 // CHECK: phi [[S_FLOAT_TY]]* 449 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* % 450 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[ARRS_PRIV]] 451 // CHECK: br i1 [[DONE]], 452 // CHECK: call void @llvm.stackrestore(i8* 453 454 // CHECK: ret void 455 456 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 457 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 458 // ... 459 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 460 // *(Type<n>-1*)rhs[<n>-1]); 461 // } 462 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 463 // arr_rhs = (int*)rhs[0]; 464 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 465 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]], 466 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32* 467 // arr_lhs = (int*)lhs[0]; 468 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 469 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]], 470 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32* 471 472 // arr_size = (size_t)lhs[1]; 473 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 474 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]], 475 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64 476 477 // arrs_rhs = (S<float>*)rhs[2]; 478 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 479 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]], 480 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]* 481 // arrs_lhs = (S<float>*)lhs[2]; 482 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 483 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]], 484 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]* 485 486 // arrs_size = (size_t)lhs[3]; 487 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 488 // CHECK: [[ARRS_SIZE_VOID:%.+]] = load i8*, i8** [[ARRS_SIZE_REF]], 489 // CHECK: [[ARRS_SIZE:%.+]] = ptrtoint i8* [[ARRS_SIZE_VOID]] to i64 490 491 // arr_lhs[:] += arr_rhs[:]; 492 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]] 493 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]] 494 // CHECK: br i1 [[ISEMPTY]], 495 // CHECK: phi i32* 496 // CHECK: [[ADD:%.+]] = mul nsw i32 555, % 497 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 498 // CHECK: br i1 [[DONE]], 499 500 // arrs_lhs = arrs_lhs.operator &(arrs_rhs); 501 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 502 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 503 // CHECK: br i1 [[ISEMPTY]], 504 // CHECK: phi [[S_FLOAT_TY]]* 505 // CHECK: call void @_Z3redR6BaseS1RKS_(% 506 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 507 // CHECK: br i1 [[DONE]], 508 509 // CHECK: ret void 510 511 // 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) %{{.+}}) 512 513 // CHECK: [[ARRS_PRIV:%.+]] = alloca [10 x [4 x [[S_FLOAT_TY]]]], 514 515 // Reduction list for runtime. 516 // CHECK: [[RED_LIST:%.+]] = alloca [3 x i8*], 517 518 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 519 520 // CHECK: [[ARR_SIZE:%.+]] = udiv exact i64 521 // CHECK: call i8* @llvm.stacksave() 522 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]], 523 524 // Check initialization of private copy. 525 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]] 526 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]] 527 // CHECK: br i1 [[ISEMPTY]], 528 // CHECK: phi i32* 529 // CHECK: store i32 888, i32* % 530 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 531 // CHECK: br i1 [[DONE]], 532 533 // Check initialization of private copy. 534 // 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 0 535 // CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]* 536 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40 537 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[BEGIN]], [[END]] 538 // CHECK: br i1 [[ISEMPTY]], 539 // CHECK: phi [[S_FLOAT_TY]]* 540 // CHECK: call void @_Z4initR6BaseS1RKS_(% 541 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 542 // CHECK: br i1 [[DONE]], 543 // CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]* 544 // CHECK: [[ARRS_PRIV_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]] to [[S_FLOAT_TY]]* 545 546 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 547 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 548 // CHECK: call void @__kmpc_for_static_init_4( 549 // Skip checks for internal operations. 550 // CHECK: call void @__kmpc_for_static_fini( 551 552 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 553 554 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 0 555 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8* 556 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]], 557 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 1 558 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8* 559 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]], 560 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 2 561 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV_BEGIN]] to i8* 562 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]], 563 564 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 565 566 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 567 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 568 // CHECK: [[BITCAST:%.+]] = bitcast [3 x i8*]* [[RED_LIST]] to i8* 569 // 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]]) 570 571 // switch(res) 572 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 573 // CHECK: i32 1, label %[[CASE1:.+]] 574 // CHECK: i32 2, label %[[CASE2:.+]] 575 // CHECK: ] 576 577 // case 1: 578 // CHECK: [[CASE1]] 579 580 // arr[:] += arr_reduction[:]; 581 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0:%.+]], i64 [[ARR_SIZE]] 582 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 583 // CHECK: br i1 [[ISEMPTY]], 584 // CHECK: phi i32* 585 // CHECK: [[ADD:%[^ ]+]] = mul nsw i32 555, % 586 // CHECK: store i32 [[ADD]], i32* % 587 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 588 // CHECK: br i1 [[DONE]], 589 590 // arrs[:] = var.operator &(arrs_reduction[:]); 591 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40 592 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]] 593 // CHECK: br i1 [[ISEMPTY]], 594 // CHECK: phi [[S_FLOAT_TY]]* 595 // CHECK: call void @_Z3redR6BaseS1RKS_(% 596 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 597 // CHECK: br i1 [[DONE]], 598 599 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 600 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 601 602 // break; 603 // CHECK: br label %[[RED_DONE]] 604 605 // case 2: 606 // CHECK: [[CASE2]] 607 608 // arr[:] += arr_reduction[:]; 609 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 610 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 611 // CHECK: br i1 [[ISEMPTY]], 612 // CHECK: phi i32* 613 // CHECK: call void @__kmpc_critical( 614 // CHECK: [[ADD:%.+]] = mul nsw i32 555, % 615 // CHECK: call void @__kmpc_end_critical( 616 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 617 // CHECK: br i1 [[DONE]], 618 619 // arrs[:] = var.operator &(arrs_reduction[:]); 620 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40 621 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]] 622 // CHECK: br i1 [[ISEMPTY]], 623 // CHECK: phi [[S_FLOAT_TY]]* 624 // CHECK: call void @__kmpc_critical( 625 // CHECK: call void @_Z3redR6BaseS1RKS_(% 626 // CHECK: call void @__kmpc_end_critical( 627 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 628 // CHECK: br i1 [[DONE]], 629 630 // break; 631 // CHECK: br label %[[RED_DONE]] 632 // CHECK: [[RED_DONE]] 633 634 // Check destruction of private copy. 635 // 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 0 636 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40 637 // CHECK: br 638 // CHECK: phi [[S_FLOAT_TY]]* 639 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* % 640 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[BEGIN]] 641 // CHECK: br i1 [[DONE]], 642 // CHECK: call void @llvm.stackrestore(i8* 643 // CHECK: call void @__kmpc_barrier( 644 645 // CHECK: ret void 646 647 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 648 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 649 // ... 650 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 651 // *(Type<n>-1*)rhs[<n>-1]); 652 // } 653 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 654 // arr_rhs = (int*)rhs[0]; 655 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 656 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]], 657 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32* 658 // arr_lhs = (int*)lhs[0]; 659 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 660 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]], 661 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32* 662 663 // arr_size = (size_t)lhs[1]; 664 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 665 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]], 666 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64 667 668 // arrs_rhs = (S<float>*)rhs[2]; 669 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 670 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]], 671 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]* 672 // arrs_lhs = (S<float>*)lhs[2]; 673 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 674 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]], 675 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]* 676 677 // arr_lhs[:] += arr_rhs[:]; 678 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]] 679 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]] 680 // CHECK: br i1 [[ISEMPTY]], 681 // CHECK: phi i32* 682 // CHECK: [[ADD:%.+]] = mul nsw i32 555, % 683 // CHECK: store i32 [[ADD]], i32* % 684 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 685 // CHECK: br i1 [[DONE]], 686 687 // arrs_lhs = arrs_lhs.operator &(arrs_rhs); 688 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 40 689 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 690 // CHECK: br i1 [[ISEMPTY]], 691 // CHECK: phi [[S_FLOAT_TY]]* 692 // CHECK: call void @_Z3redR6BaseS1RKS_(% 693 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 694 // CHECK: br i1 [[DONE]], 695 696 // CHECK: ret void 697 698 // CHECK: define internal void [[MAIN_MICROTASK3]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** dereferenceable(8) %{{.+}}) 699 700 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***, 701 702 // Reduction list for runtime. 703 // CHECK: [[RED_LIST:%.+]] = alloca [2 x i8*], 704 705 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 706 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]], 707 708 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64 709 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW:%.+]] to i64 710 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]] 711 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64) 712 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1 713 // CHECK: call i8* @llvm.stacksave() 714 // CHECK: [[VAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]], 715 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 716 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], 717 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 718 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 719 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 720 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64) 721 // CHECK: [[PSEUDO_VAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR2_PRIV]], i64 [[OFFSET]] 722 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** % 723 // CHECK: store [[S_FLOAT_TY]]* [[PSEUDO_VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]] 724 // CHECK: ret void 725 726 // CHECK: define internal void [[MAIN_MICROTASK4]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [5 x [[S_FLOAT_TY]]]* dereferenceable(60) %{{.+}}) 727 728 // CHECK: [[VVAR2_ORIG_ADDR:%.+]] = alloca [5 x [[S_FLOAT_TY]]]*, 729 // CHECK: [[VVAR2_PRIV:%.+]] = alloca [5 x [[S_FLOAT_TY]]], 730 731 // Reduction list for runtime. 732 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 733 734 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 735 // CHECK: [[VVAR2_ORIG:%.+]] = load [5 x [[S_FLOAT_TY]]]*, [5 x [[S_FLOAT_TY]]]** [[VVAR2_ORIG_ADDR]], 736 737 // CHECK: [[LOW:%.+]] = getelementptr inbounds [5 x [[S_FLOAT_TY]]], [5 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]], i64 0, i64 0 738 // CHECK: [[ORIG_START:%.+]] = bitcast [5 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]] to [[S_FLOAT_TY]]* 739 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 740 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 741 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 742 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64) 743 // CHECK: [[VVAR2_PRIV_PTR:%.+]] = bitcast [5 x [[S_FLOAT_TY]]]* [[VVAR2_PRIV]] to [[S_FLOAT_TY]]* 744 // CHECK: [[VVAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VVAR2_PRIV_PTR]], i64 [[OFFSET]] 745 // CHECK: ret void 746 747 // CHECK: define internal void [[MAIN_MICROTASK5]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* dereferenceable(48) %{{.+}}) 748 749 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 750 // CHECK: [[VAR3_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]], 751 752 // Reduction list for runtime. 753 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 754 755 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 756 757 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 758 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 759 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 760 761 // CHECK: [[LOW:%.+]] = getelementptr inbounds [4 x [[S_FLOAT_TY]]], [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], i64 0, i64 1 762 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 763 764 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 765 // CHECK: [[ORIG_START:%.+]] = bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 766 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 767 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 768 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 769 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64) 770 // CHECK: [[VAR3_PRIV_PTR:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]] to [[S_FLOAT_TY]]* 771 // CHECK: [[PSEUDO_VAR3_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR3_PRIV_PTR]], i64 [[OFFSET]] 772 // CHECK: [[VAR3_PRIV:%.+]] = bitcast [[S_FLOAT_TY]]* [[PSEUDO_VAR3_PRIV]] to [4 x [[S_FLOAT_TY]]]* 773 774 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 775 776 // CHECK: ret void 777 778 // CHECK: define internal void [[MAIN_MICROTASK6]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* dereferenceable(48) %{{.+}}) 779 780 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 781 // CHECK: [[VAR3_PRIV:%.+]] = alloca [4 x [[S_FLOAT_TY]]], 782 783 // Reduction list for runtime. 784 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 785 786 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 787 788 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 789 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 790 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 791 // CHECK: getelementptr inbounds [4 x [[S_FLOAT_TY]]], [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], i32 0, i32 0 792 // CHECK: bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 793 // CHECK: getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 4 794 795 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 796 797 // CHECK: ret void 798 799 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT_42]]() 800 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]], 801 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]]) 802 // 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 803 // Not interested in this one: 804 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4, 805 // 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]+}}*, [42 x [[S_INT_TY]]]*, [2 x i32]*, i32*, [2 x [[S_INT_TY]]]*, [[S_INT_TY]]*)* [[TMAIN_MICROTASK2:@.+]] to void 806 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]* 807 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]* 808 // CHECK: ret 809 // 810 // 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) %{{.+}}) 811 // CHECK: alloca i{{[0-9]+}}, 812 // CHECK: alloca i{{[0-9]+}}, 813 // CHECK: alloca i{{[0-9]+}}, 814 // CHECK: alloca i{{[0-9]+}}, 815 // CHECK: alloca i{{[0-9]+}}, 816 // CHECK: alloca i{{[0-9]+}}, 817 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 818 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], 819 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_INT_TY]], 820 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca i{{[0-9]+}}, 821 822 // Reduction list for runtime. 823 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 824 825 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 826 827 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 828 // CHECK: [[VAR1_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 829 // CHECK: [[T_VAR1_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 830 831 // For + reduction operation initial value of private variable is 0. 832 // CHECK: store i32 321, i32* % 833 834 // For & reduction operation initial value of private variable is ones in all bits. 835 // CHECK: call void @_Z4initR6BaseS1RKS_( 836 837 // For && reduction operation initial value of private variable is 1.0. 838 // CHECK: call void @_Z5init2R6BaseS1RKS_( 839 840 // For min reduction operation initial value of private variable is largest repesentable value. 841 // CHECK: sdiv i32 432, % 842 843 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 844 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 845 // CHECK: call void @__kmpc_for_static_init_4( 846 // Skip checks for internal operations. 847 // CHECK: call void @__kmpc_for_static_fini( 848 849 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 850 851 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 852 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR_PRIV]] to i8* 853 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 854 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 855 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR_PRIV]] to i8* 856 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 857 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 858 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_PRIV]] to i8* 859 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 860 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 861 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR1_PRIV]] to i8* 862 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 863 864 // res = __kmpc_reduce_nowait(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 865 866 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 867 // 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]]) 868 869 // switch(res) 870 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 871 // CHECK: i32 1, label %[[CASE1:.+]] 872 // CHECK: i32 2, label %[[CASE2:.+]] 873 // CHECK: ] 874 875 // case 1: 876 // t_var += t_var_reduction; 877 // CHECK: add nsw i32 1513, % 878 879 // var = var.operator &(var_reduction); 880 // CHECK: call void @_Z3redR6BaseS1RKS_(% 881 882 // var1 = var1.operator &&(var1_reduction); 883 // CHECK: mul nsw i32 17, % 884 885 // t_var1 = min(t_var1, t_var1_reduction); 886 // CHECK: sub nsw i32 47, % 887 888 // __kmpc_end_reduce_nowait(<loc>, <gtid>, &<lock>); 889 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 890 891 // break; 892 // CHECK: br label %[[RED_DONE]] 893 894 // case 2: 895 // t_var += t_var_reduction; 896 // CHECK: call void @__kmpc_critical( 897 // CHECK: add nsw i32 1513, % 898 // CHECK: call void @__kmpc_end_critical( 899 900 // var = var.operator &(var_reduction); 901 // CHECK: call void @__kmpc_critical( 902 // CHECK: call void @_Z3redR6BaseS1RKS_(% 903 // CHECK: call void @__kmpc_end_critical( 904 905 // var1 = var1.operator &&(var1_reduction); 906 // CHECK: call void @__kmpc_critical( 907 // CHECK: mul nsw i32 17, % 908 // CHECK: call void @__kmpc_end_critical( 909 910 // t_var1 = min(t_var1, t_var1_reduction); 911 // CHECK: call void @__kmpc_critical( 912 // CHECK: sub nsw i32 47, % 913 // CHECK: call void @__kmpc_end_critical( 914 915 // break; 916 // CHECK: br label %[[RED_DONE]] 917 // CHECK: [[RED_DONE]] 918 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) 919 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* 920 // CHECK: ret void 921 922 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 923 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 924 // ... 925 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 926 // *(Type<n>-1*)rhs[<n>-1]); 927 // } 928 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 929 // t_var_lhs = (i{{[0-9]+}}*)lhs[0]; 930 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 931 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 932 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to i{{[0-9]+}}* 933 // t_var_rhs = (i{{[0-9]+}}*)rhs[0]; 934 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 935 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 936 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to i{{[0-9]+}}* 937 938 // var_lhs = (S<i{{[0-9]+}}>*)lhs[1]; 939 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 940 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 941 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_INT_TY]]* 942 // var_rhs = (S<i{{[0-9]+}}>*)rhs[1]; 943 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 944 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 945 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_INT_TY]]* 946 947 // var1_lhs = (S<i{{[0-9]+}}>*)lhs[2]; 948 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 949 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 950 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_INT_TY]]* 951 // var1_rhs = (S<i{{[0-9]+}}>*)rhs[2]; 952 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 953 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 954 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_INT_TY]]* 955 956 // t_var1_lhs = (i{{[0-9]+}}*)lhs[3]; 957 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 958 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 959 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to i{{[0-9]+}}* 960 // t_var1_rhs = (i{{[0-9]+}}*)rhs[3]; 961 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 962 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 963 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to i{{[0-9]+}}* 964 965 // t_var_lhs += t_var_rhs; 966 // CHECK: add nsw i32 1513, % 967 968 // var_lhs = var_lhs.operator &(var_rhs); 969 // CHECK: call void @_Z3redR6BaseS1RKS_(% 970 971 // var1_lhs = var1_lhs.operator &&(var1_rhs); 972 // CHECK: mul nsw i32 17, % 973 974 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 975 // CHECK: sub nsw i32 47, % 976 // CHECK: ret void 977 978 // CHECK: define internal void [[TMAIN_MICROTASK2]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [42 x [[S_INT_TY]]]* dereferenceable(504) %{{.*}}, [2 x i32]* dereferenceable(8) %{{.*}}, i32* dereferenceable(4) %{{.*}}, [2 x [[S_INT_TY]]]* dereferenceable(24) %{{.*}}, [[S_INT_TY]]* dereferenceable(12) %{{.*}}) 979 980 // CHECK: [[ARR_ORIG_ADDR:%.+]] = alloca [42 x [[S_INT_TY]]]*, 981 // CHECK: [[ARR_PRIV:%.+]] = alloca [40 x [[S_INT_TY]]], 982 983 // Reduction list for runtime. 984 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 985 986 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 987 988 // CHECK: [[ARR_ORIG:%.+]] = load [42 x [[S_INT_TY]]]*, [42 x [[S_INT_TY]]]** [[ARR_ORIG_ADDR]], 989 // CHECK: [[LOW:%.+]] = getelementptr inbounds [42 x [[S_INT_TY]]], [42 x [[S_INT_TY]]]* [[ARR_ORIG]], i64 0, i64 1 990 // CHECK: [[ORIG_START:%.+]] = bitcast [42 x [[S_INT_TY]]]* [[ARR_ORIG]] to [[S_INT_TY]]* 991 // CHECK: [[START:%.+]] = ptrtoint [[S_INT_TY]]* [[ORIG_START]] to i64 992 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_INT_TY]]* [[LOW]] to i64 993 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 994 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_INT_TY]]* getelementptr ([[S_INT_TY]], [[S_INT_TY]]* null, i32 1) to i64) 995 // CHECK: [[ARR_PRIV_PTR:%.+]] = bitcast [40 x [[S_INT_TY]]]* [[ARR_PRIV]] to [[S_INT_TY]]* 996 // CHECK: [[PSEUDO_ARR_PRIV:%.+]] = getelementptr [[S_INT_TY]], [[S_INT_TY]]* [[ARR_PRIV_PTR]], i64 [[OFFSET]] 997 // CHECK: [[ARR_PRIV:%.+]] = bitcast [[S_INT_TY]]* [[PSEUDO_ARR_PRIV]] to [42 x [[S_INT_TY]]]* 998 999 // CHECK: ret void 1000 1001 #endif 1002 1003