1 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap %s 2 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s 3 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap %s 4 5 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap --check-prefix SIMD-ONLY0 %s 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 -allow-deprecated-dag-overlap --check-prefix SIMD-ONLY0 %s 8 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}} 9 // expected-no-diagnostics 10 #ifndef HEADER 11 #define HEADER 12 13 enum omp_allocator_handle_t { 14 omp_null_allocator = 0, 15 omp_default_mem_alloc = 1, 16 omp_large_cap_mem_alloc = 2, 17 omp_const_mem_alloc = 3, 18 omp_high_bw_mem_alloc = 4, 19 omp_low_lat_mem_alloc = 5, 20 omp_cgroup_mem_alloc = 6, 21 omp_pteam_mem_alloc = 7, 22 omp_thread_mem_alloc = 8, 23 KMP_ALLOCATOR_MAX_HANDLE = __UINTPTR_MAX__ 24 }; 25 26 volatile double g, g_orig; 27 volatile double &g1 = g_orig; 28 29 struct BaseS { 30 int x; 31 }; 32 struct BaseS1 { 33 float y; 34 }; 35 36 template <class T> 37 struct S : public BaseS, public BaseS1 { 38 T f; 39 S(T a) : f(a + g) {} 40 S() : f(g) {} 41 ~S() {} 42 }; 43 void red(BaseS1&, const BaseS1&); 44 void red_plus(BaseS1&, const BaseS1&); 45 void init(BaseS1&, const BaseS1&); 46 void init1(BaseS1&, const BaseS1&); 47 void init2(BaseS1&, const BaseS1&); 48 void init_plus(BaseS1&, const BaseS1&); 49 #pragma omp declare reduction(operator& : BaseS1 : red(omp_out, omp_in)) initializer(init(omp_priv, omp_orig)) 50 #pragma omp declare reduction(+ : BaseS1 : red_plus(omp_out, omp_in)) initializer(init_plus(omp_priv, omp_orig)) 51 #pragma omp declare reduction(&& : S<float>, S<int> : omp_out.f *= omp_in.f) initializer(init1(omp_priv, omp_orig)) 52 53 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { %{{[^,]+}}, %{{[^,]+}}, float } 54 // CHECK-DAG: [[S_INT_TY:%.+]] = type { %{{[^,]+}}, %{{[^,]+}}, i{{[0-9]+}} } 55 // CHECK-DAG: [[ATOMIC_REDUCE_BARRIER_LOC:@.+]] = private unnamed_addr global %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 56 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr global %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8* 57 // CHECK-DAG: [[REDUCTION_LOC:@.+]] = private unnamed_addr global %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 58 // CHECK-DAG: [[REDUCTION_LOCK:@.+]] = common global [8 x i32] zeroinitializer 59 60 #pragma omp declare reduction(operator&& : int : omp_out = 111 & omp_in) 61 template <typename T, int length> 62 T tmain() { 63 T t; 64 S<T> test; 65 T t_var = T(), t_var1; 66 T vec[] = {1, 2}; 67 S<T> s_arr[] = {1, 2}; 68 S<T> &var = test; 69 S<T> var1; 70 S<T> arr[length]; 71 #pragma omp declare reduction(operator& : T : omp_out = 15 + omp_in) 72 #pragma omp declare reduction(operator+ : T : omp_out = 1513 + omp_in) initializer(omp_priv = 321) 73 #pragma omp declare reduction(min : T : omp_out = 47 - omp_in) initializer(omp_priv = 432 / omp_orig) 74 #pragma omp declare reduction(operator&& : S<T> : omp_out.f = 17 * omp_in.f) initializer(init2(omp_priv, omp_orig)) 75 #pragma omp declare reduction(operator&& : T : omp_out = 17 * omp_in) 76 #pragma omp parallel 77 #pragma omp for reduction(+ : t_var) reduction(& : var) reduction(&& : var1) reduction(min : t_var1) nowait 78 for (int i = 0; i < 2; ++i) { 79 vec[i] = t_var; 80 s_arr[i] = var; 81 } 82 #pragma omp parallel 83 #pragma omp for reduction(&& : t_var) 84 for (int i = 0; i < 2; ++i) { 85 vec[i] = t_var; 86 s_arr[i] = var; 87 } 88 #pragma omp parallel 89 #pragma omp for reduction(+ : arr[1:length-2]) 90 for (int i = 0; i < 2; ++i) { 91 vec[i] = t_var; 92 s_arr[i] = var; 93 } 94 return T(); 95 } 96 97 extern S<float> **foo(); 98 99 #pragma omp declare reduction(operator- : float, double : omp_out = 333 + omp_in) 100 #pragma omp declare reduction(min : float, double : omp_out = 555 + omp_in) 101 int main() { 102 #pragma omp declare reduction(operator+ : float, double : omp_out = 222 - omp_in) initializer(omp_priv = -1) 103 S<float> test; 104 float t_var = 0, t_var1; 105 int vec[] = {1, 2}; 106 S<float> s_arr[] = {1, 2, 3, 4}; 107 S<float> &var = test; 108 S<float> var1, arrs[10][4]; 109 S<float> **var2 = foo(); 110 S<float> vvar2[5]; 111 S<float>(&var3)[4] = s_arr; 112 #pragma omp declare reduction(operator+ : int : omp_out = 555 * omp_in) initializer(omp_priv = 888) 113 #pragma omp parallel 114 #pragma omp for reduction(+ : t_var) reduction(& : var) reduction(&& : var1) reduction(min : t_var1) 115 for (int i = 0; i < 2; ++i) { 116 vec[i] = t_var; 117 s_arr[i] = var; 118 } 119 int arr[10][vec[1]]; 120 #pragma omp parallel for reduction(+ : arr[1][ : vec[1]]) reduction(& : arrs[1 : vec[1]][1 : 2]) 121 for (int i = 0; i < 10; ++i) 122 ++arr[1][i]; 123 #pragma omp parallel 124 #pragma omp for reduction(+ : arr) reduction(& : arrs) 125 for (int i = 0; i < 10; ++i) 126 ++arr[1][i]; 127 #pragma omp parallel 128 #pragma omp for reduction(& : var2[0 : 5][1 : 6]) 129 for (int i = 0; i < 10; ++i) 130 ; 131 #pragma omp parallel 132 #pragma omp for reduction(& : vvar2[0 : 5]) 133 for (int i = 0; i < 10; ++i) 134 ; 135 #pragma omp parallel 136 #pragma omp for reduction(& : var3[1 : 2]) 137 for (int i = 0; i < 10; ++i) 138 ; 139 #pragma omp parallel 140 #pragma omp for reduction(& : var3) allocate(omp_cgroup_mem_alloc: var3) 141 for (int i = 0; i < 10; ++i) 142 ; 143 return tmain<int, 42>(); 144 } 145 146 // CHECK: define {{.*}}i{{[0-9]+}} @main() 147 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]], 148 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]]) 149 // 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 150 // 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 151 // 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 152 // 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 153 // 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 154 // 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 155 // 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 156 // CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT_42:@.+]]() 157 // CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]* 158 // CHECK: ret 159 // 160 // 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) %{{.+}}) 161 // CHECK: [[T_VAR_PRIV:%.+]] = alloca float, 162 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 163 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 164 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca float, 165 166 // Reduction list for runtime. 167 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 168 169 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 170 171 // CHECK: [[T_VAR_REF:%.+]] = load float*, float** % 172 // CHECK: [[VAR1_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** % 173 // CHECK: [[T_VAR1_REF:%.+]] = load float*, float** % 174 175 // For + reduction operation initial value of private variable is -1. 176 // CHECK: store float -1.0{{.+}}, float* 177 178 // For & reduction operation initial value of private variable is defined by call of 'init()' function. 179 // CHECK: call {{.*}}void @_Z4initR6BaseS1RKS_( 180 181 // For && reduction operation initial value of private variable is 1.0. 182 // CHECK: call {{.*}}void @_Z5init1R6BaseS1RKS_( 183 184 // For min reduction operation initial value of private variable is largest repesentable value. 185 // CHECK: [[INIT:%.+]] = load float, float* @ 186 // CHECK: store float [[INIT]], float* [[T_VAR1_PRIV]], 187 188 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 189 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 190 // CHECK: call void @__kmpc_for_static_init_4( 191 // Skip checks for internal operations. 192 // CHECK: call void @__kmpc_for_static_fini( 193 194 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 195 196 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 197 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR_PRIV]] to i8* 198 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 199 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 200 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_PRIV]] to i8* 201 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 202 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 203 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_PRIV]] to i8* 204 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 205 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 206 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR1_PRIV]] to i8* 207 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 208 209 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 210 211 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 212 // 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]]) 213 214 // switch(res) 215 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 216 // CHECK: i32 1, label %[[CASE1:.+]] 217 // CHECK: i32 2, label %[[CASE2:.+]] 218 // CHECK: ] 219 220 // case 1: 221 // t_var += t_var_reduction; 222 // CHECK: fsub float 2.220000e+02, % 223 224 // var = var.operator &(var_reduction); 225 // CHECK: call {{.*}}void @_Z3redR6BaseS1RKS_( 226 227 // var1 = var1.operator &&(var1_reduction); 228 // CHECK: fmul float 229 230 // t_var1 = min(t_var1, t_var1_reduction); 231 // CHECK: fadd float 5.550000e+02, % 232 233 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 234 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 235 236 // break; 237 // CHECK: br label %[[RED_DONE]] 238 239 // case 2: 240 // t_var += t_var_reduction; 241 // CHECK: call void @__kmpc_critical( 242 // CHECK: fsub float 2.220000e+02, % 243 // CHECK: call void @__kmpc_end_critical( 244 245 // var = var.operator &(var_reduction); 246 // CHECK: call void @__kmpc_critical( 247 // CHECK: call {{.*}}void @_Z3redR6BaseS1RKS_( 248 // CHECK: call void @__kmpc_end_critical( 249 250 // var1 = var1.operator &&(var1_reduction); 251 // CHECK: call void @__kmpc_critical( 252 // CHECK: fmul float 253 // CHECK: call void @__kmpc_end_critical( 254 255 // t_var1 = min(t_var1, t_var1_reduction); 256 // CHECK: call void @__kmpc_critical( 257 // CHECK: fadd float 5.550000e+02, % 258 // CHECK: call void @__kmpc_end_critical( 259 260 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 261 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 262 263 // break; 264 // CHECK: br label %[[RED_DONE]] 265 // CHECK: [[RED_DONE]] 266 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 267 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* 268 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 269 270 // CHECK: ret void 271 272 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 273 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 274 // ... 275 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 276 // *(Type<n>-1*)rhs[<n>-1]); 277 // } 278 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 279 // t_var_lhs = (float*)lhs[0]; 280 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 281 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 282 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to float* 283 // t_var_rhs = (float*)rhs[0]; 284 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 285 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 286 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to float* 287 288 // var_lhs = (S<float>*)lhs[1]; 289 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 290 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 291 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_FLOAT_TY]]* 292 // var_rhs = (S<float>*)rhs[1]; 293 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 294 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 295 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_FLOAT_TY]]* 296 297 // var1_lhs = (S<float>*)lhs[2]; 298 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 299 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 300 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_FLOAT_TY]]* 301 // var1_rhs = (S<float>*)rhs[2]; 302 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 303 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 304 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_FLOAT_TY]]* 305 306 // t_var1_lhs = (float*)lhs[3]; 307 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 308 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 309 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to float* 310 // t_var1_rhs = (float*)rhs[3]; 311 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 312 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 313 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to float* 314 315 // t_var_lhs += t_var_rhs; 316 // CHECK: fsub float 2.220000e+02, % 317 318 // var_lhs = var_lhs.operator &(var_rhs); 319 // CHECK: call {{.*}}void @_Z3redR6BaseS1RKS_( 320 321 // var1_lhs = var1_lhs.operator &&(var1_rhs); 322 // CHECK: fmul float 323 324 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 325 // CHECK: fadd float 5.550000e+02, % 326 // CHECK: ret void 327 328 // 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) %{{.+}}) 329 330 // Reduction list for runtime. 331 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 332 333 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 334 // CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}** 335 // CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}** [[ARR_ADDR:%.+]], 336 // CHECK: [[ARR:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[ARR_ADDR]], 337 338 // CHECK: [[UB_CAST:%.+]] = ptrtoint i32* [[UB1_UP:%.+]] to i64 339 // CHECK: [[LB_CAST:%.+]] = ptrtoint i32* [[LB1_0:%.+]] to i64 340 // CHECK: [[DIFF:%.+]] = sub i64 [[UB_CAST]], [[LB_CAST]] 341 // CHECK: [[SIZE_1:%.+]] = sdiv exact i64 [[DIFF]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64) 342 // CHECK: [[ARR_SIZE:%.+]] = add nuw i64 [[SIZE_1]], 1 343 // CHECK: call i8* @llvm.stacksave() 344 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]], 345 346 // Check initialization of private copy. 347 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]] 348 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]] 349 // CHECK: br i1 [[ISEMPTY]], 350 // CHECK: phi i32* 351 // CHECK: store i32 888, i32* % 352 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 353 // CHECK: br i1 [[DONE]], 354 355 // CHECK: [[ARRS_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[ARRS_SIZE:%.+]], 356 357 // Check initialization of private copy. 358 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]] 359 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]] 360 // CHECK: br i1 [[ISEMPTY]], 361 // CHECK: phi [[S_FLOAT_TY]]* 362 // CHECK: call void @_Z4initR6BaseS1RKS_(% 363 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 364 // CHECK: br i1 [[DONE]], 365 366 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 367 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 368 // CHECK: call void @__kmpc_for_static_init_4( 369 // Skip checks for internal operations. 370 // CHECK: call void @__kmpc_for_static_fini( 371 372 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 373 374 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 375 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8* 376 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]], 377 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 378 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8* 379 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]], 380 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 381 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV]] to i8* 382 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]], 383 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 384 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARRS_SIZE]] to i8* 385 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_SIZE_REF]], 386 387 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 388 389 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 390 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 391 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 392 // 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]]) 393 394 // switch(res) 395 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 396 // CHECK: i32 1, label %[[CASE1:.+]] 397 // CHECK: i32 2, label %[[CASE2:.+]] 398 // CHECK: ] 399 400 // case 1: 401 // CHECK: [[CASE1]] 402 403 // arr[:] += arr_reduction[:]; 404 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 405 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 406 // CHECK: br i1 [[ISEMPTY]], 407 // CHECK: phi i32* 408 // CHECK: [[ADD:%.+]] = mul nsw i32 555, % 409 // CHECK: store i32 [[ADD]], i32* % 410 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 411 // CHECK: br i1 [[DONE]], 412 413 // arrs[:] = var.operator &(arrs_reduction[:]); 414 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 415 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 416 // CHECK: br i1 [[ISEMPTY]], 417 // CHECK: phi [[S_FLOAT_TY]]* 418 // CHECK: call void @_Z3redR6BaseS1RKS_(% 419 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 420 // CHECK: br i1 [[DONE]], 421 422 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 423 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 424 425 // break; 426 // CHECK: br label %[[RED_DONE]] 427 428 // case 2: 429 // CHECK: [[CASE2]] 430 431 // arr[:] += arr_reduction[:]; 432 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 433 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 434 // CHECK: br i1 [[ISEMPTY]], 435 // CHECK: phi i32* 436 // CHECK: call void @__kmpc_critical( 437 // CHECK: [[ADD:%.+]] = mul nsw i32 555, % 438 // CHECK: call void @__kmpc_end_critical( 439 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 440 // CHECK: br i1 [[DONE]], 441 442 // arrs[:] = var.operator &(arrs_reduction[:]); 443 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 444 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 445 // CHECK: br i1 [[ISEMPTY]], 446 // CHECK: phi [[S_FLOAT_TY]]* 447 // CHECK: call void @__kmpc_critical( 448 // CHECK: call void @_Z3redR6BaseS1RKS_(% 449 // CHECK: call void @__kmpc_end_critical( 450 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 451 // CHECK: br i1 [[DONE]], 452 453 // break; 454 // CHECK: br label %[[RED_DONE]] 455 // CHECK: [[RED_DONE]] 456 457 // Check destruction of private copy. 458 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]] 459 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]] 460 // CHECK: br i1 [[ISEMPTY]], 461 // CHECK: phi [[S_FLOAT_TY]]* 462 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* % 463 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[ARRS_PRIV]] 464 // CHECK: br i1 [[DONE]], 465 // CHECK: call void @llvm.stackrestore(i8* 466 467 // CHECK: ret void 468 469 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 470 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 471 // ... 472 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 473 // *(Type<n>-1*)rhs[<n>-1]); 474 // } 475 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 476 // arr_rhs = (int*)rhs[0]; 477 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 478 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]], 479 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32* 480 // arr_lhs = (int*)lhs[0]; 481 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 482 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]], 483 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32* 484 485 // arr_size = (size_t)lhs[1]; 486 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 487 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]], 488 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64 489 490 // arrs_rhs = (S<float>*)rhs[2]; 491 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 492 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]], 493 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]* 494 // arrs_lhs = (S<float>*)lhs[2]; 495 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 496 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]], 497 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]* 498 499 // arrs_size = (size_t)lhs[3]; 500 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 501 // CHECK: [[ARRS_SIZE_VOID:%.+]] = load i8*, i8** [[ARRS_SIZE_REF]], 502 // CHECK: [[ARRS_SIZE:%.+]] = ptrtoint i8* [[ARRS_SIZE_VOID]] to i64 503 504 // arr_lhs[:] += arr_rhs[:]; 505 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]] 506 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]] 507 // CHECK: br i1 [[ISEMPTY]], 508 // CHECK: phi i32* 509 // CHECK: [[ADD:%.+]] = mul nsw i32 555, % 510 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 511 // CHECK: br i1 [[DONE]], 512 513 // arrs_lhs = arrs_lhs.operator &(arrs_rhs); 514 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 515 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 516 // CHECK: br i1 [[ISEMPTY]], 517 // CHECK: phi [[S_FLOAT_TY]]* 518 // CHECK: call void @_Z3redR6BaseS1RKS_(% 519 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 520 // CHECK: br i1 [[DONE]], 521 522 // CHECK: ret void 523 524 // 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) %{{.+}}) 525 526 // CHECK: [[ARRS_PRIV:%.+]] = alloca [10 x [4 x [[S_FLOAT_TY]]]], 527 528 // Reduction list for runtime. 529 // CHECK: [[RED_LIST:%.+]] = alloca [3 x i8*], 530 531 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 532 533 // CHECK: [[ARR_SIZE:%.+]] = udiv exact i64 534 // CHECK: call i8* @llvm.stacksave() 535 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]], 536 537 // Check initialization of private copy. 538 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]] 539 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]] 540 // CHECK: br i1 [[ISEMPTY]], 541 // CHECK: phi i32* 542 // CHECK: store i32 888, i32* % 543 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 544 // CHECK: br i1 [[DONE]], 545 546 // Check initialization of private copy. 547 // 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 0 548 // CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]* 549 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40 550 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[BEGIN]], [[END]] 551 // CHECK: br i1 [[ISEMPTY]], 552 // CHECK: phi [[S_FLOAT_TY]]* 553 // CHECK: call void @_Z4initR6BaseS1RKS_(% 554 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 555 // CHECK: br i1 [[DONE]], 556 // CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]* 557 // CHECK: [[ARRS_PRIV_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]] to [[S_FLOAT_TY]]* 558 559 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 560 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 561 // CHECK: call void @__kmpc_for_static_init_4( 562 // Skip checks for internal operations. 563 // CHECK: call void @__kmpc_for_static_fini( 564 565 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 566 567 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 0 568 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8* 569 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]], 570 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 1 571 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8* 572 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]], 573 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 2 574 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV_BEGIN]] to i8* 575 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]], 576 577 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 578 579 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 580 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 581 // CHECK: [[BITCAST:%.+]] = bitcast [3 x i8*]* [[RED_LIST]] to i8* 582 // 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]]) 583 584 // switch(res) 585 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 586 // CHECK: i32 1, label %[[CASE1:.+]] 587 // CHECK: i32 2, label %[[CASE2:.+]] 588 // CHECK: ] 589 590 // case 1: 591 // CHECK: [[CASE1]] 592 593 // arr[:] += arr_reduction[:]; 594 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0:%.+]], i64 [[ARR_SIZE]] 595 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 596 // CHECK: br i1 [[ISEMPTY]], 597 // CHECK: phi i32* 598 // CHECK: [[ADD:%[^ ]+]] = mul nsw i32 555, % 599 // CHECK: store i32 [[ADD]], i32* % 600 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 601 // CHECK: br i1 [[DONE]], 602 603 // arrs[:] = var.operator &(arrs_reduction[:]); 604 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40 605 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]] 606 // CHECK: br i1 [[ISEMPTY]], 607 // CHECK: phi [[S_FLOAT_TY]]* 608 // CHECK: call void @_Z3redR6BaseS1RKS_(% 609 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 610 // CHECK: br i1 [[DONE]], 611 612 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 613 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 614 615 // break; 616 // CHECK: br label %[[RED_DONE]] 617 618 // case 2: 619 // CHECK: [[CASE2]] 620 621 // arr[:] += arr_reduction[:]; 622 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 623 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 624 // CHECK: br i1 [[ISEMPTY]], 625 // CHECK: phi i32* 626 // CHECK: call void @__kmpc_critical( 627 // CHECK: [[ADD:%.+]] = mul nsw i32 555, % 628 // CHECK: call void @__kmpc_end_critical( 629 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 630 // CHECK: br i1 [[DONE]], 631 632 // arrs[:] = var.operator &(arrs_reduction[:]); 633 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40 634 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]] 635 // CHECK: br i1 [[ISEMPTY]], 636 // CHECK: phi [[S_FLOAT_TY]]* 637 // CHECK: call void @__kmpc_critical( 638 // CHECK: call void @_Z3redR6BaseS1RKS_(% 639 // CHECK: call void @__kmpc_end_critical( 640 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 641 // CHECK: br i1 [[DONE]], 642 643 // break; 644 // CHECK: br label %[[RED_DONE]] 645 // CHECK: [[RED_DONE]] 646 647 // Check destruction of private copy. 648 // 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 0 649 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40 650 // CHECK: br 651 // CHECK: phi [[S_FLOAT_TY]]* 652 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* % 653 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[BEGIN]] 654 // CHECK: br i1 [[DONE]], 655 // CHECK: call void @llvm.stackrestore(i8* 656 // CHECK: call void @__kmpc_barrier( 657 658 // CHECK: ret void 659 660 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 661 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 662 // ... 663 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 664 // *(Type<n>-1*)rhs[<n>-1]); 665 // } 666 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 667 // arr_rhs = (int*)rhs[0]; 668 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 669 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]], 670 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32* 671 // arr_lhs = (int*)lhs[0]; 672 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 673 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]], 674 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32* 675 676 // arr_size = (size_t)lhs[1]; 677 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 678 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]], 679 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64 680 681 // arrs_rhs = (S<float>*)rhs[2]; 682 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 683 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]], 684 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]* 685 // arrs_lhs = (S<float>*)lhs[2]; 686 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 687 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]], 688 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]* 689 690 // arr_lhs[:] += arr_rhs[:]; 691 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]] 692 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]] 693 // CHECK: br i1 [[ISEMPTY]], 694 // CHECK: phi i32* 695 // CHECK: [[ADD:%.+]] = mul nsw i32 555, % 696 // CHECK: store i32 [[ADD]], i32* % 697 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 698 // CHECK: br i1 [[DONE]], 699 700 // arrs_lhs = arrs_lhs.operator &(arrs_rhs); 701 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 40 702 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 703 // CHECK: br i1 [[ISEMPTY]], 704 // CHECK: phi [[S_FLOAT_TY]]* 705 // CHECK: call void @_Z3redR6BaseS1RKS_(% 706 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 707 // CHECK: br i1 [[DONE]], 708 709 // CHECK: ret void 710 711 // CHECK: define internal void [[MAIN_MICROTASK3]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** dereferenceable(8) %{{.+}}) 712 713 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***, 714 715 // Reduction list for runtime. 716 // CHECK: [[RED_LIST:%.+]] = alloca [2 x i8*], 717 718 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 719 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]], 720 721 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64 722 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW:%.+]] to i64 723 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]] 724 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64) 725 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1 726 // CHECK: call i8* @llvm.stacksave() 727 // CHECK: [[VAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]], 728 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 729 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], 730 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 731 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 732 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 733 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64) 734 // CHECK: [[PSEUDO_VAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR2_PRIV]], i64 [[OFFSET]] 735 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** % 736 // CHECK: store [[S_FLOAT_TY]]* [[PSEUDO_VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]] 737 // CHECK: ret void 738 739 // CHECK: define internal void [[MAIN_MICROTASK4]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [5 x [[S_FLOAT_TY]]]* dereferenceable(60) %{{.+}}) 740 741 // CHECK: [[VVAR2_ORIG_ADDR:%.+]] = alloca [5 x [[S_FLOAT_TY]]]*, 742 // CHECK: [[VVAR2_PRIV:%.+]] = alloca [5 x [[S_FLOAT_TY]]], 743 744 // Reduction list for runtime. 745 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 746 747 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 748 // CHECK: [[VVAR2_ORIG:%.+]] = load [5 x [[S_FLOAT_TY]]]*, [5 x [[S_FLOAT_TY]]]** [[VVAR2_ORIG_ADDR]], 749 750 // CHECK: [[LOW:%.+]] = getelementptr inbounds [5 x [[S_FLOAT_TY]]], [5 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]], i64 0, i64 0 751 // CHECK: [[ORIG_START:%.+]] = bitcast [5 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]] to [[S_FLOAT_TY]]* 752 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 753 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 754 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 755 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64) 756 // CHECK: [[VVAR2_PRIV_PTR:%.+]] = bitcast [5 x [[S_FLOAT_TY]]]* [[VVAR2_PRIV]] to [[S_FLOAT_TY]]* 757 // CHECK: [[VVAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VVAR2_PRIV_PTR]], i64 [[OFFSET]] 758 // CHECK: ret void 759 760 // CHECK: define internal void [[MAIN_MICROTASK5]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* dereferenceable(48) %{{.+}}) 761 762 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 763 // CHECK: [[VAR3_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]], 764 765 // Reduction list for runtime. 766 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 767 768 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 769 770 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 771 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 772 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 773 774 // CHECK: [[LOW:%.+]] = getelementptr inbounds [4 x [[S_FLOAT_TY]]], [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], i64 0, i64 1 775 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 776 777 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 778 // CHECK: [[ORIG_START:%.+]] = bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 779 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 780 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 781 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 782 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64) 783 // CHECK: [[VAR3_PRIV_PTR:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]] to [[S_FLOAT_TY]]* 784 // CHECK: [[PSEUDO_VAR3_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR3_PRIV_PTR]], i64 [[OFFSET]] 785 // CHECK: [[VAR3_PRIV:%.+]] = bitcast [[S_FLOAT_TY]]* [[PSEUDO_VAR3_PRIV]] to [4 x [[S_FLOAT_TY]]]* 786 787 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 788 789 // CHECK: ret void 790 791 // CHECK: define internal void [[MAIN_MICROTASK6]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* dereferenceable(48) %{{.+}}) 792 793 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 794 795 // Reduction list for runtime. 796 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 797 798 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 799 800 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 801 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 802 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 803 // CHECK: [[VAR3_VOID_PTR:%.+]] = call i8* @__kmpc_alloc(i32 [[GTID:%.+]], i64 48, i8* inttoptr (i64 6 to i8*)) 804 // CHECK: [[VAR3_PRIV:%.+]] = bitcast i8* [[VAR3_VOID_PTR]] to [4 x %struct.S]* 805 // CHECK: getelementptr inbounds [4 x [[S_FLOAT_TY]]], [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], i32 0, i32 0 806 // CHECK: bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 807 // CHECK: getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 4 808 809 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 810 // CHECK: call void @__kmpc_free(i32 [[GTID]], i8* [[VAR3_VOID_PTR]], i8* inttoptr (i64 6 to i8*)) 811 // CHECK: ret void 812 813 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT_42]]() 814 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]], 815 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]]) 816 // 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 817 // Not interested in this one: 818 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4, 819 // 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 820 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]* 821 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]* 822 // CHECK: ret 823 // 824 // 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) %{{.+}}) 825 // CHECK: alloca i{{[0-9]+}}, 826 // CHECK: alloca i{{[0-9]+}}, 827 // CHECK: alloca i{{[0-9]+}}, 828 // CHECK: alloca i{{[0-9]+}}, 829 // CHECK: alloca i{{[0-9]+}}, 830 // CHECK: alloca i{{[0-9]+}}, 831 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 832 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], 833 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_INT_TY]], 834 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca i{{[0-9]+}}, 835 836 // Reduction list for runtime. 837 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 838 839 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 840 841 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 842 // CHECK: [[VAR1_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 843 // CHECK: [[T_VAR1_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 844 845 // For + reduction operation initial value of private variable is 0. 846 // CHECK: store i32 321, i32* % 847 848 // For & reduction operation initial value of private variable is ones in all bits. 849 // CHECK: call void @_Z4initR6BaseS1RKS_( 850 851 // For && reduction operation initial value of private variable is 1.0. 852 // CHECK: call void @_Z5init2R6BaseS1RKS_( 853 854 // For min reduction operation initial value of private variable is largest repesentable value. 855 // CHECK: sdiv i32 432, % 856 857 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 858 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 859 // CHECK: call void @__kmpc_for_static_init_4( 860 // Skip checks for internal operations. 861 // CHECK: call void @__kmpc_for_static_fini( 862 863 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 864 865 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 866 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR_PRIV]] to i8* 867 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 868 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 869 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR_PRIV]] to i8* 870 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 871 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 872 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_PRIV]] to i8* 873 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 874 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 875 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR1_PRIV]] to i8* 876 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 877 878 // res = __kmpc_reduce_nowait(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 879 880 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 881 // 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]]) 882 883 // switch(res) 884 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 885 // CHECK: i32 1, label %[[CASE1:.+]] 886 // CHECK: i32 2, label %[[CASE2:.+]] 887 // CHECK: ] 888 889 // case 1: 890 // t_var += t_var_reduction; 891 // CHECK: add nsw i32 1513, % 892 893 // var = var.operator &(var_reduction); 894 // CHECK: call void @_Z3redR6BaseS1RKS_(% 895 896 // var1 = var1.operator &&(var1_reduction); 897 // CHECK: mul nsw i32 17, % 898 899 // t_var1 = min(t_var1, t_var1_reduction); 900 // CHECK: sub nsw i32 47, % 901 902 // __kmpc_end_reduce_nowait(<loc>, <gtid>, &<lock>); 903 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 904 905 // break; 906 // CHECK: br label %[[RED_DONE]] 907 908 // case 2: 909 // t_var += t_var_reduction; 910 // CHECK: call void @__kmpc_critical( 911 // CHECK: add nsw i32 1513, % 912 // CHECK: call void @__kmpc_end_critical( 913 914 // var = var.operator &(var_reduction); 915 // CHECK: call void @__kmpc_critical( 916 // CHECK: call void @_Z3redR6BaseS1RKS_(% 917 // CHECK: call void @__kmpc_end_critical( 918 919 // var1 = var1.operator &&(var1_reduction); 920 // CHECK: call void @__kmpc_critical( 921 // CHECK: mul nsw i32 17, % 922 // CHECK: call void @__kmpc_end_critical( 923 924 // t_var1 = min(t_var1, t_var1_reduction); 925 // CHECK: call void @__kmpc_critical( 926 // CHECK: sub nsw i32 47, % 927 // CHECK: call void @__kmpc_end_critical( 928 929 // break; 930 // CHECK: br label %[[RED_DONE]] 931 // CHECK: [[RED_DONE]] 932 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) 933 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* 934 // CHECK: ret void 935 936 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 937 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 938 // ... 939 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 940 // *(Type<n>-1*)rhs[<n>-1]); 941 // } 942 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 943 // t_var_lhs = (i{{[0-9]+}}*)lhs[0]; 944 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 945 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 946 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to i{{[0-9]+}}* 947 // t_var_rhs = (i{{[0-9]+}}*)rhs[0]; 948 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 949 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 950 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to i{{[0-9]+}}* 951 952 // var_lhs = (S<i{{[0-9]+}}>*)lhs[1]; 953 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 954 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 955 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_INT_TY]]* 956 // var_rhs = (S<i{{[0-9]+}}>*)rhs[1]; 957 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 958 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 959 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_INT_TY]]* 960 961 // var1_lhs = (S<i{{[0-9]+}}>*)lhs[2]; 962 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 963 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 964 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_INT_TY]]* 965 // var1_rhs = (S<i{{[0-9]+}}>*)rhs[2]; 966 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 967 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 968 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_INT_TY]]* 969 970 // t_var1_lhs = (i{{[0-9]+}}*)lhs[3]; 971 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 972 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 973 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to i{{[0-9]+}}* 974 // t_var1_rhs = (i{{[0-9]+}}*)rhs[3]; 975 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 976 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 977 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to i{{[0-9]+}}* 978 979 // t_var_lhs += t_var_rhs; 980 // CHECK: add nsw i32 1513, % 981 982 // var_lhs = var_lhs.operator &(var_rhs); 983 // CHECK: call void @_Z3redR6BaseS1RKS_(% 984 985 // var1_lhs = var1_lhs.operator &&(var1_rhs); 986 // CHECK: mul nsw i32 17, % 987 988 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 989 // CHECK: sub nsw i32 47, % 990 // CHECK: ret void 991 992 // 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) %{{.*}}) 993 994 // CHECK: [[ARR_ORIG_ADDR:%.+]] = alloca [42 x [[S_INT_TY]]]*, 995 // CHECK: [[ARR_PRIV:%.+]] = alloca [40 x [[S_INT_TY]]], 996 997 // Reduction list for runtime. 998 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 999 1000 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1001 1002 // CHECK: [[ARR_ORIG:%.+]] = load [42 x [[S_INT_TY]]]*, [42 x [[S_INT_TY]]]** [[ARR_ORIG_ADDR]], 1003 // CHECK: [[LOW:%.+]] = getelementptr inbounds [42 x [[S_INT_TY]]], [42 x [[S_INT_TY]]]* [[ARR_ORIG]], i64 0, i64 1 1004 // CHECK: [[ORIG_START:%.+]] = bitcast [42 x [[S_INT_TY]]]* [[ARR_ORIG]] to [[S_INT_TY]]* 1005 // CHECK: [[START:%.+]] = ptrtoint [[S_INT_TY]]* [[ORIG_START]] to i64 1006 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_INT_TY]]* [[LOW]] to i64 1007 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1008 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_INT_TY]]* getelementptr ([[S_INT_TY]], [[S_INT_TY]]* null, i32 1) to i64) 1009 // CHECK: [[ARR_PRIV_PTR:%.+]] = bitcast [40 x [[S_INT_TY]]]* [[ARR_PRIV]] to [[S_INT_TY]]* 1010 // CHECK: [[PSEUDO_ARR_PRIV:%.+]] = getelementptr [[S_INT_TY]], [[S_INT_TY]]* [[ARR_PRIV_PTR]], i64 [[OFFSET]] 1011 // CHECK: [[ARR_PRIV:%.+]] = bitcast [[S_INT_TY]]* [[PSEUDO_ARR_PRIV]] to [42 x [[S_INT_TY]]]* 1012 1013 // CHECK: ret void 1014 1015 #endif 1016 1017