1 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap %s 2 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s 3 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap %s 4 // RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DLAMBDA -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap -check-prefix=LAMBDA %s 5 // RUN: %clang_cc1 -verify -fopenmp -x c++ -fblocks -DBLOCKS -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap -check-prefix=BLOCKS %s 6 7 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap --check-prefix SIMD-ONLY0 %s 8 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s 9 // RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap --check-prefix SIMD-ONLY0 %s 10 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -std=c++11 -DLAMBDA -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap --check-prefix SIMD-ONLY0 %s 11 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -fblocks -DBLOCKS -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap --check-prefix SIMD-ONLY0 %s 12 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}} 13 // expected-no-diagnostics 14 #ifndef HEADER 15 #define HEADER 16 17 volatile double g, g_orig; 18 volatile double &g1 = g_orig; 19 20 template <class T> 21 struct S { 22 T f; 23 S(T a) : f(a + g) {} 24 S() : f(g) {} 25 operator T() { return T(); } 26 S &operator&(const S &) { return *this; } 27 ~S() {} 28 }; 29 30 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { float } 31 // CHECK-DAG: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} } 32 // CHECK-DAG: [[ATOMIC_REDUCE_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 33 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8* 34 // CHECK-DAG: [[REDUCTION_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 35 // CHECK-DAG: [[REDUCTION_LOCK:@.+]] = common global [8 x i32] zeroinitializer 36 37 template <typename T, int length> 38 T tmain() { 39 T t; 40 S<T> test; 41 T t_var = T(), t_var1; 42 T vec[] = {1, 2}; 43 S<T> s_arr[] = {1, 2}; 44 S<T> &var = test; 45 S<T> var1; 46 S<T> arr[length]; 47 #pragma omp parallel 48 #pragma omp for reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) nowait 49 for (int i = 0; i < 2; ++i) { 50 vec[i] = t_var; 51 s_arr[i] = var; 52 } 53 #pragma omp parallel 54 #pragma omp for reduction(&& : t_var) 55 for (int i = 0; i < 2; ++i) { 56 vec[i] = t_var; 57 s_arr[i] = var; 58 } 59 #pragma omp parallel 60 #pragma omp for reduction(+ : arr[1:length-2]) 61 for (int i = 0; i < 2; ++i) { 62 vec[i] = t_var; 63 s_arr[i] = var; 64 } 65 return T(); 66 } 67 68 extern S<float> **foo(); 69 70 int main() { 71 #ifdef LAMBDA 72 // LAMBDA: [[G:@.+]] = global double 73 // LAMBDA-LABEL: @main 74 // LAMBDA: call void [[OUTER_LAMBDA:@.+]]( 75 [&]() { 76 // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]]( 77 // LAMBDA: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}) 78 #pragma omp parallel 79 #pragma omp for reduction(+:g, g1) 80 for (int i = 0; i < 2; ++i) { 81 // LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}) 82 // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca double, 83 84 // Reduction list for runtime. 85 // LAMBDA: [[RED_LIST:%.+]] = alloca [2 x i8*], 86 87 // LAMBDA: store double 0.0{{.+}}, double* [[G_PRIVATE_ADDR]] 88 // LAMBDA: call void @__kmpc_for_static_init_4( 89 g = 1; 90 g1 = 1; 91 // LAMBDA: store double 1.0{{.+}}, double* [[G_PRIVATE_ADDR]], 92 // LAMBDA: [[G_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 93 // LAMBDA: store double* [[G_PRIVATE_ADDR]], double** [[G_PRIVATE_ADDR_REF]] 94 // LAMBDA: call void [[INNER_LAMBDA:@.+]](%{{.+}}* [[ARG]]) 95 // LAMBDA: call void @__kmpc_for_static_fini( 96 97 // LAMBDA: [[G_PRIV_REF:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[RED_LIST]], i64 0, i64 0 98 // LAMBDA: [[BITCAST:%.+]] = bitcast double* [[G_PRIVATE_ADDR]] to i8* 99 // LAMBDA: store i8* [[BITCAST]], i8** [[G_PRIV_REF]], 100 // LAMBDA: call i32 @__kmpc_reduce( 101 // LAMBDA: switch i32 %{{.+}}, label %[[REDUCTION_DONE:.+]] [ 102 // LAMBDA: i32 1, label %[[CASE1:.+]] 103 // LAMBDA: i32 2, label %[[CASE2:.+]] 104 // LAMBDA: [[CASE1]] 105 // LAMBDA: [[G_VAL:%.+]] = load double, double* [[G]] 106 // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 107 // LAMBDA: [[ADD:%.+]] = fadd double [[G_VAL]], [[G_PRIV_VAL]] 108 // LAMBDA: store double [[ADD]], double* [[G]] 109 // LAMBDA: call void @__kmpc_end_reduce( 110 // LAMBDA: br label %[[REDUCTION_DONE]] 111 // LAMBDA: [[CASE2]] 112 // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 113 // LAMBDA: fadd double 114 // LAMBDA: cmpxchg i64* 115 // LAMBDA: call void @__kmpc_end_reduce( 116 // LAMBDA: br label %[[REDUCTION_DONE]] 117 // LAMBDA: [[REDUCTION_DONE]] 118 // LAMBDA: ret void 119 [&]() { 120 // LAMBDA: define {{.+}} void [[INNER_LAMBDA]](%{{.+}}* [[ARG_PTR:%.+]]) 121 // LAMBDA: store %{{.+}}* [[ARG_PTR]], %{{.+}}** [[ARG_PTR_REF:%.+]], 122 g = 2; 123 g1 = 2; 124 // LAMBDA: [[ARG_PTR:%.+]] = load %{{.+}}*, %{{.+}}** [[ARG_PTR_REF]] 125 // LAMBDA: [[G_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 126 // LAMBDA: [[G_REF:%.+]] = load double*, double** [[G_PTR_REF]] 127 // LAMBDA: store double 2.0{{.+}}, double* [[G_REF]] 128 }(); 129 } 130 }(); 131 return 0; 132 #elif defined(BLOCKS) 133 // BLOCKS: [[G:@.+]] = global double 134 // BLOCKS-LABEL: @main 135 // BLOCKS: call void {{%.+}}(i8 136 ^{ 137 // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8* 138 // BLOCKS: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}) 139 #pragma omp parallel 140 #pragma omp for reduction(-:g, g1) 141 for (int i = 0; i < 2; ++i) { 142 // BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}) 143 // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca double, 144 145 // Reduction list for runtime. 146 // BLOCKS: [[RED_LIST:%.+]] = alloca [2 x i8*], 147 148 // BLOCKS: store double 0.0{{.+}}, double* [[G_PRIVATE_ADDR]] 149 g = 1; 150 g1 = 1; 151 // BLOCKS: call void @__kmpc_for_static_init_4( 152 // BLOCKS: store double 1.0{{.+}}, double* [[G_PRIVATE_ADDR]], 153 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 154 // BLOCKS: double* [[G_PRIVATE_ADDR]] 155 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 156 // BLOCKS: call void {{%.+}}(i8 157 // BLOCKS: call void @__kmpc_for_static_fini( 158 159 // BLOCKS: [[G_PRIV_REF:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[RED_LIST]], i64 0, i64 0 160 // BLOCKS: [[BITCAST:%.+]] = bitcast double* [[G_PRIVATE_ADDR]] to i8* 161 // BLOCKS: store i8* [[BITCAST]], i8** [[G_PRIV_REF]], 162 // BLOCKS: call i32 @__kmpc_reduce( 163 // BLOCKS: switch i32 %{{.+}}, label %[[REDUCTION_DONE:.+]] [ 164 // BLOCKS: i32 1, label %[[CASE1:.+]] 165 // BLOCKS: i32 2, label %[[CASE2:.+]] 166 // BLOCKS: [[CASE1]] 167 // BLOCKS: [[G_VAL:%.+]] = load double, double* [[G]] 168 // BLOCKS: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 169 // BLOCKS: [[ADD:%.+]] = fadd double [[G_VAL]], [[G_PRIV_VAL]] 170 // BLOCKS: store double [[ADD]], double* [[G]] 171 // BLOCKS: call void @__kmpc_end_reduce( 172 // BLOCKS: br label %[[REDUCTION_DONE]] 173 // BLOCKS: [[CASE2]] 174 // BLOCKS: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 175 // BLOCKS: fadd double 176 // BLOCKS: cmpxchg i64* 177 // BLOCKS: call void @__kmpc_end_reduce( 178 // BLOCKS: br label %[[REDUCTION_DONE]] 179 // BLOCKS: [[REDUCTION_DONE]] 180 // BLOCKS: ret void 181 ^{ 182 // BLOCKS: define {{.+}} void {{@.+}}(i8* 183 g = 2; 184 g1 = 2; 185 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 186 // BLOCKS: store double 2.0{{.+}}, double* 187 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 188 // BLOCKS: ret 189 }(); 190 } 191 }(); 192 return 0; 193 #else 194 S<float> test; 195 float t_var = 0, t_var1; 196 int vec[] = {1, 2}; 197 S<float> s_arr[] = {1, 2, 3, 4}; 198 S<float> &var = test; 199 S<float> var1, arrs[10][4]; 200 S<float> **var2 = foo(); 201 S<float> vvar2[5]; 202 S<float> (&var3)[4] = s_arr; 203 #pragma omp parallel 204 #pragma omp for reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) 205 for (int i = 0; i < 2; ++i) { 206 vec[i] = t_var; 207 s_arr[i] = var; 208 } 209 int arr[10][vec[1]]; 210 #pragma omp parallel for reduction(+:arr[1][:vec[1]]) reduction(&:arrs[1:vec[1]][1:2]) 211 for (int i = 0; i < 10; ++i) 212 ++arr[1][i]; 213 #pragma omp parallel 214 #pragma omp for reduction(+:arr) reduction(&:arrs) 215 for (int i = 0; i < 10; ++i) 216 ++arr[1][i]; 217 // arr is a VLA, but the array section has constant length so we can generate a constant sized array! 218 #pragma omp parallel 219 #pragma omp for reduction(+:arr[1][0:2]) 220 for (int i = 0; i < 10; ++i) 221 ++arr[1][i]; 222 #pragma omp parallel 223 #pragma omp for reduction(& : var2[0 : 5][1 : 6]) 224 for (int i = 0; i < 10; ++i) 225 ; 226 #pragma omp parallel 227 #pragma omp for reduction(& : var2[1][1 : 6]) 228 for (int i = 0; i < 10; ++i) 229 ; 230 #pragma omp parallel 231 #pragma omp for reduction(& : var2[1 : 1][1 : 6]) 232 for (int i = 0; i < 10; ++i) 233 ; 234 #pragma omp parallel 235 #pragma omp for reduction(& : var2[1 : 1][1]) 236 for (int i = 0; i < 10; ++i) 237 ; 238 #pragma omp parallel 239 #pragma omp for reduction(& : vvar2[0 : 5]) 240 for (int i = 0; i < 10; ++i) 241 ; 242 #pragma omp parallel 243 #pragma omp for reduction(& : var3[1 : 2]) 244 for (int i = 0; i < 10; ++i) 245 ; 246 #pragma omp parallel 247 #pragma omp for reduction(& : var3[ : 2]) 248 for (int i = 0; i < 10; ++i) 249 ; 250 // TODO: The compiler should also be able to generate a constant sized array in this case! 251 #pragma omp parallel 252 #pragma omp for reduction(& : var3[2 : ]) 253 for (int i = 0; i < 10; ++i) 254 ; 255 #pragma omp parallel 256 #pragma omp for reduction(& : var3) 257 for (int i = 0; i < 10; ++i) 258 ; 259 return tmain<int, 42>(); 260 #endif 261 } 262 263 // CHECK: define {{.*}}i{{[0-9]+}} @main() 264 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]], 265 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]]) 266 // 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 267 // 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 268 // 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 269 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 3, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, i64, i64, i32*)* [[MAIN_MICROTASK3:@.+]] to void 270 // 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_MICROTASK4:@.+]] to void 271 // 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_MICROTASK5:@.+]] to void 272 // 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_MICROTASK6:@.+]] to void 273 // 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_MICROTASK7:@.+]] to void 274 // 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_MICROTASK8:@.+]] to void 275 // 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_MICROTASK9:@.+]] to void 276 // 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_MICROTASK10:@.+]] to void 277 // 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_MICROTASK11:@.+]] to void 278 // 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_MICROTASK12:@.+]] to void 279 // CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT_42:@.+]]() 280 // CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]* 281 // CHECK: ret 282 // 283 // CHECK: define internal void [[MAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, float* nonnull align 4 dereferenceable(4) %{{.+}}, [[S_FLOAT_TY]]* nonnull align 4 dereferenceable(4) %{{.+}}, [[S_FLOAT_TY]]* nonnull align 4 dereferenceable(4) %{{.+}}, float* nonnull align 4 dereferenceable(4) %{{.+}}, [2 x i32]* nonnull align 4 dereferenceable(8) %vec, [4 x [[S_FLOAT_TY]]]* nonnull align 4 dereferenceable(16) %{{.+}}) 284 // CHECK: [[T_VAR_PRIV:%.+]] = alloca float, 285 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 286 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 287 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca float, 288 289 // Reduction list for runtime. 290 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 291 292 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 293 294 // CHECK: [[T_VAR_REF:%.+]] = load float*, float** % 295 // CHECK: [[VAR_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** % 296 // CHECK: [[VAR1_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** % 297 // CHECK: [[T_VAR1_REF:%.+]] = load float*, float** % 298 299 // For + reduction operation initial value of private variable is 0. 300 // CHECK: store float 0.0{{.+}}, float* [[T_VAR_PRIV]], 301 302 // CHECK: [[VAR_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** % 303 304 // For & reduction operation initial value of private variable is ones in all bits. 305 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 306 307 // For && reduction operation initial value of private variable is 1.0. 308 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[VAR1_PRIV]]) 309 310 // For min reduction operation initial value of private variable is largest repesentable value. 311 // CHECK: store float 0x47EFFFFFE0000000, float* [[T_VAR1_PRIV]], 312 313 314 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 315 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 316 // CHECK: call void @__kmpc_for_static_init_4( 317 // Skip checks for internal operations. 318 // CHECK: call void @__kmpc_for_static_fini( 319 320 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 321 322 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 323 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR_PRIV]] to i8* 324 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 325 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 326 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_PRIV]] to i8* 327 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 328 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 329 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_PRIV]] to i8* 330 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 331 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 332 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR1_PRIV]] to i8* 333 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 334 335 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 336 337 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 338 // 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]]) 339 340 // switch(res) 341 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 342 // CHECK: i32 1, label %[[CASE1:.+]] 343 // CHECK: i32 2, label %[[CASE2:.+]] 344 // CHECK: ] 345 346 // case 1: 347 // t_var += t_var_reduction; 348 // CHECK: [[T_VAR_VAL:%.+]] = load float, float* [[T_VAR_REF]], 349 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load float, float* [[T_VAR_PRIV]], 350 // CHECK: [[UP:%.+]] = fadd float [[T_VAR_VAL]], [[T_VAR_PRIV_VAL]] 351 // CHECK: store float [[UP]], float* [[T_VAR_REF]], 352 353 // var = var.operator &(var_reduction); 354 // CHECK: [[UP:%.+]] = call nonnull align 4 dereferenceable(4) [[S_FLOAT_TY]]* @{{.+}}([[S_FLOAT_TY]]* [[VAR_REF]], [[S_FLOAT_TY]]* nonnull align 4 dereferenceable(4) [[VAR_PRIV]]) 355 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_REF]] to i8* 356 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[UP]] to i8* 357 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 358 359 // var1 = var1.operator &&(var1_reduction); 360 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_REF]]) 361 // CHECK: [[VAR1_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 362 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 363 // CHECK: [[TRUE]] 364 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_PRIV]]) 365 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 366 // CHECK: br label %[[END2]] 367 // CHECK: [[END2]] 368 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 369 // CHECK: [[CONV:%.+]] = uitofp i1 [[COND_LVALUE]] to float 370 // CHECK: call void @{{.+}}([[S_FLOAT_TY]]* [[COND_LVALUE:%.+]], float [[CONV]]) 371 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_REF]] to i8* 372 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[COND_LVALUE]] to i8* 373 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 374 375 // t_var1 = min(t_var1, t_var1_reduction); 376 // CHECK: [[T_VAR1_VAL:%.+]] = load float, float* [[T_VAR1_REF]], 377 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load float, float* [[T_VAR1_PRIV]], 378 // CHECK: [[CMP:%.+]] = fcmp olt float [[T_VAR1_VAL]], [[T_VAR1_PRIV_VAL]] 379 // CHECK: br i1 [[CMP]] 380 // CHECK: [[UP:%.+]] = phi float 381 // CHECK: store float [[UP]], float* [[T_VAR1_REF]], 382 383 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 384 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 385 386 // break; 387 // CHECK: br label %[[RED_DONE]] 388 389 // case 2: 390 // t_var += t_var_reduction; 391 // CHECK: load float, float* [[T_VAR_PRIV]] 392 // CHECK: [[T_VAR_REF_INT:%.+]] = bitcast float* [[T_VAR_REF]] to i32* 393 // CHECK: [[OLD1:%.+]] = load atomic i32, i32* [[T_VAR_REF_INT]] monotonic, 394 // CHECK: br label %[[CONT:.+]] 395 // CHECK: [[CONT]] 396 // CHECK: [[ORIG_OLD_INT:%.+]] = phi i32 [ [[OLD1]], %{{.+}} ], [ [[OLD2:%.+]], %[[CONT]] ] 397 // CHECK: fadd float 398 // CHECK: [[UP_INT:%.+]] = load i32, i32* 399 // CHECK: [[T_VAR_REF_INT:%.+]] = bitcast float* [[T_VAR_REF]] to i32* 400 // CHECK: [[RES:%.+]] = cmpxchg i32* [[T_VAR_REF_INT]], i32 [[ORIG_OLD_INT]], i32 [[UP_INT]] monotonic monotonic 401 // CHECK: [[OLD2:%.+]] = extractvalue { i32, i1 } [[RES]], 0 402 // CHECK: [[SUCCESS_FAIL:%.+]] = extractvalue { i32, i1 } [[RES]], 1 403 // CHECK: br i1 [[SUCCESS_FAIL]], label %[[ATOMIC_DONE:.+]], label %[[CONT]] 404 // CHECK: [[ATOMIC_DONE]] 405 406 // var = var.operator &(var_reduction); 407 // CHECK: call void @__kmpc_critical( 408 // CHECK: [[UP:%.+]] = call nonnull align 4 dereferenceable(4) [[S_FLOAT_TY]]* @{{.+}}([[S_FLOAT_TY]]* [[VAR_REF]], [[S_FLOAT_TY]]* nonnull align 4 dereferenceable(4) [[VAR_PRIV]]) 409 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_REF]] to i8* 410 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[UP]] to i8* 411 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 412 // CHECK: call void @__kmpc_end_critical( 413 414 // var1 = var1.operator &&(var1_reduction); 415 // CHECK: call void @__kmpc_critical( 416 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_REF]]) 417 // CHECK: [[VAR1_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 418 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 419 // CHECK: [[TRUE]] 420 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_PRIV]]) 421 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 422 // CHECK: br label %[[END2]] 423 // CHECK: [[END2]] 424 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 425 // CHECK: [[CONV:%.+]] = uitofp i1 [[COND_LVALUE]] to float 426 // CHECK: call void @{{.+}}([[S_FLOAT_TY]]* [[COND_LVALUE:%.+]], float [[CONV]]) 427 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_REF]] to i8* 428 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[COND_LVALUE]] to i8* 429 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 430 // CHECK: call void @__kmpc_end_critical( 431 432 // t_var1 = min(t_var1, t_var1_reduction); 433 // CHECK: load float, float* [[T_VAR1_PRIV]] 434 // CHECK: [[T_VAR1_REF_INT:%.+]] = bitcast float* [[T_VAR1_REF]] to i32* 435 // CHECK: [[OLD1:%.+]] = load atomic i32, i32* [[T_VAR1_REF_INT]] monotonic, 436 // CHECK: br label %[[CONT:.+]] 437 // CHECK: [[CONT]] 438 // CHECK: [[ORIG_OLD_INT:%.+]] = phi i32 [ [[OLD1]], %{{.+}} ], [ [[OLD2:%.+]], %{{.+}} ] 439 // CHECK: [[CMP:%.+]] = fcmp olt float 440 // CHECK: br i1 [[CMP]] 441 // CHECK: phi float 442 // CHECK: [[UP_INT:%.+]] = load i32 443 // CHECK: [[T_VAR1_REF_INT:%.+]] = bitcast float* [[T_VAR1_REF]] to i32* 444 // CHECK: [[RES:%.+]] = cmpxchg i32* [[T_VAR1_REF_INT]], i32 [[ORIG_OLD_INT]], i32 [[UP_INT]] monotonic monotonic 445 // CHECK: [[OLD2:%.+]] = extractvalue { i32, i1 } [[RES]], 0 446 // CHECK: [[SUCCESS_FAIL:%.+]] = extractvalue { i32, i1 } [[RES]], 1 447 // CHECK: br i1 [[SUCCESS_FAIL]], label %[[ATOMIC_DONE:.+]], label %[[CONT]] 448 // CHECK: [[ATOMIC_DONE]] 449 450 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 451 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 452 453 // break; 454 // CHECK: br label %[[RED_DONE]] 455 // CHECK: [[RED_DONE]] 456 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 457 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* 458 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 459 460 // CHECK: ret void 461 462 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 463 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 464 // ... 465 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 466 // *(Type<n>-1*)rhs[<n>-1]); 467 // } 468 // CHECK: define internal void [[REDUCTION_FUNC]](i8* %0, i8* %1) 469 // t_var_lhs = (float*)lhs[0]; 470 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 471 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 472 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to float* 473 // t_var_rhs = (float*)rhs[0]; 474 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 475 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 476 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to float* 477 478 // var_lhs = (S<float>*)lhs[1]; 479 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 480 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 481 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_FLOAT_TY]]* 482 // var_rhs = (S<float>*)rhs[1]; 483 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 484 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 485 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_FLOAT_TY]]* 486 487 // var1_lhs = (S<float>*)lhs[2]; 488 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 489 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 490 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_FLOAT_TY]]* 491 // var1_rhs = (S<float>*)rhs[2]; 492 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 493 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 494 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_FLOAT_TY]]* 495 496 // t_var1_lhs = (float*)lhs[3]; 497 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 498 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 499 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to float* 500 // t_var1_rhs = (float*)rhs[3]; 501 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 502 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 503 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to float* 504 505 // t_var_lhs += t_var_rhs; 506 // CHECK: [[T_VAR_LHS_VAL:%.+]] = load float, float* [[T_VAR_LHS]], 507 // CHECK: [[T_VAR_RHS_VAL:%.+]] = load float, float* [[T_VAR_RHS]], 508 // CHECK: [[UP:%.+]] = fadd float [[T_VAR_LHS_VAL]], [[T_VAR_RHS_VAL]] 509 // CHECK: store float [[UP]], float* [[T_VAR_LHS]], 510 511 // var_lhs = var_lhs.operator &(var_rhs); 512 // CHECK: [[UP:%.+]] = call nonnull align 4 dereferenceable(4) [[S_FLOAT_TY]]* @{{.+}}([[S_FLOAT_TY]]* [[VAR_LHS]], [[S_FLOAT_TY]]* nonnull align 4 dereferenceable(4) [[VAR_RHS]]) 513 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_LHS]] to i8* 514 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[UP]] to i8* 515 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 516 517 // var1_lhs = var1_lhs.operator &&(var1_rhs); 518 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_LHS]]) 519 // CHECK: [[VAR1_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 520 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 521 // CHECK: [[TRUE]] 522 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_RHS]]) 523 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 524 // CHECK: br label %[[END2]] 525 // CHECK: [[END2]] 526 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 527 // CHECK: [[CONV:%.+]] = uitofp i1 [[COND_LVALUE]] to float 528 // CHECK: call void @{{.+}}([[S_FLOAT_TY]]* [[COND_LVALUE:%.+]], float [[CONV]]) 529 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_LHS]] to i8* 530 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[COND_LVALUE]] to i8* 531 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 532 533 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 534 // CHECK: [[T_VAR1_LHS_VAL:%.+]] = load float, float* [[T_VAR1_LHS]], 535 // CHECK: [[T_VAR1_RHS_VAL:%.+]] = load float, float* [[T_VAR1_RHS]], 536 // CHECK: [[CMP:%.+]] = fcmp olt float [[T_VAR1_LHS_VAL]], [[T_VAR1_RHS_VAL]] 537 // CHECK: br i1 [[CMP]] 538 // CHECK: [[UP:%.+]] = phi float 539 // CHECK: store float [[UP]], float* [[T_VAR1_LHS]], 540 // CHECK: ret void 541 542 // CHECK: define internal void [[MAIN_MICROTASK1]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i64 %{{.+}}, i64 %{{.+}}, i32* {{.+}} %{{.+}}, [2 x i32]* nonnull align 4 dereferenceable(8) %{{.+}}, [10 x [4 x [[S_FLOAT_TY]]]]* nonnull align 4 dereferenceable(160) %{{.+}}) 543 544 // Reduction list for runtime. 545 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 546 547 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 548 // CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}** 549 // CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}** [[ARR_ADDR:%.+]], 550 // CHECK: [[ARR:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[ARR_ADDR]], 551 552 // CHECK: [[UB_CAST:%.+]] = ptrtoint i32* [[UB1_UP:%.+]] to i64 553 // CHECK: [[LB_CAST:%.+]] = ptrtoint i32* [[LB1_0:%.+]] to i64 554 // CHECK: [[DIFF:%.+]] = sub i64 [[UB_CAST]], [[LB_CAST]] 555 // CHECK: [[SIZE_1:%.+]] = sdiv exact i64 [[DIFF]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64) 556 // CHECK: [[ARR_SIZE:%.+]] = add nuw i64 [[SIZE_1]], 1 557 // CHECK: call i8* @llvm.stacksave() 558 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]], 559 560 // Check initialization of private copy. 561 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]] 562 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]] 563 // CHECK: br i1 [[ISEMPTY]], 564 // CHECK: phi i32* 565 // CHECK: store i32 0, i32* % 566 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 567 // CHECK: br i1 [[DONE]], 568 569 // CHECK: [[ARRS_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[ARRS_SIZE:%.+]], 570 571 // Check initialization of private copy. 572 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]] 573 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]] 574 // CHECK: br i1 [[ISEMPTY]], 575 // CHECK: phi [[S_FLOAT_TY]]* 576 // CHECK: call void @_ZN1SIfEC1Ev([[S_FLOAT_TY]]* % 577 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 578 // CHECK: br i1 [[DONE]], 579 580 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 581 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 582 // CHECK: call void @__kmpc_for_static_init_4( 583 // Skip checks for internal operations. 584 // CHECK: call void @__kmpc_for_static_fini( 585 586 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 587 588 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 589 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8* 590 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]], 591 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 592 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8* 593 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]], 594 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 595 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV]] to i8* 596 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]], 597 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 598 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARRS_SIZE]] to i8* 599 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_SIZE_REF]], 600 601 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 602 603 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 604 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 605 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 606 // 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]]) 607 608 // switch(res) 609 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 610 // CHECK: i32 1, label %[[CASE1:.+]] 611 // CHECK: i32 2, label %[[CASE2:.+]] 612 // CHECK: ] 613 614 // case 1: 615 // CHECK: [[CASE1]] 616 617 // arr[:] += arr_reduction[:]; 618 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 619 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 620 // CHECK: br i1 [[ISEMPTY]], 621 // CHECK: phi i32* 622 // CHECK: [[ADD:%.+]] = add nsw i32 % 623 // CHECK: store i32 [[ADD]], i32* % 624 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 625 // CHECK: br i1 [[DONE]], 626 627 // arrs[:] = var.operator &(arrs_reduction[:]); 628 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 629 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 630 // CHECK: br i1 [[ISEMPTY]], 631 // CHECK: phi [[S_FLOAT_TY]]* 632 // CHECK: [[AND:%.+]] = call nonnull align 4 dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* nonnull align 4 dereferenceable(4) %{{.+}}) 633 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 634 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %{{.+}}, i8* align 4 [[BITCAST]], i64 4, i1 false) 635 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 636 // CHECK: br i1 [[DONE]], 637 638 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 639 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 640 641 // break; 642 // CHECK: br label %[[RED_DONE]] 643 644 // case 2: 645 // CHECK: [[CASE2]] 646 647 // arr[:] += arr_reduction[:]; 648 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 649 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 650 // CHECK: br i1 [[ISEMPTY]], 651 // CHECK: phi i32* 652 // CHECK: atomicrmw add i32* %{{.+}}, i32 %{{.+}} monotonic 653 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 654 // CHECK: br i1 [[DONE]], 655 656 // arrs[:] = var.operator &(arrs_reduction[:]); 657 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 658 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 659 // CHECK: br i1 [[ISEMPTY]], 660 // CHECK: phi [[S_FLOAT_TY]]* 661 // CHECK: call void @__kmpc_critical( 662 // CHECK: [[AND:%.+]] = call nonnull align 4 dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* nonnull align 4 dereferenceable(4) %{{.+}}) 663 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 664 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %{{.+}}, i8* align 4 [[BITCAST]], i64 4, i1 false) 665 // CHECK: call void @__kmpc_end_critical( 666 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 667 // CHECK: br i1 [[DONE]], 668 669 // break; 670 // CHECK: br label %[[RED_DONE]] 671 // CHECK: [[RED_DONE]] 672 673 // Check destruction of private copy. 674 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]] 675 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]] 676 // CHECK: br i1 [[ISEMPTY]], 677 // CHECK: phi [[S_FLOAT_TY]]* 678 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* % 679 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[ARRS_PRIV]] 680 // CHECK: br i1 [[DONE]], 681 // CHECK: call void @llvm.stackrestore(i8* 682 683 // CHECK: ret void 684 685 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 686 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 687 // ... 688 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 689 // *(Type<n>-1*)rhs[<n>-1]); 690 // } 691 // CHECK: define internal void [[REDUCTION_FUNC]](i8* %0, i8* %1) 692 // arr_rhs = (int*)rhs[0]; 693 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 694 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]], 695 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32* 696 // arr_lhs = (int*)lhs[0]; 697 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 698 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]], 699 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32* 700 701 // arr_size = (size_t)lhs[1]; 702 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 703 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]], 704 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64 705 706 // arrs_rhs = (S<float>*)rhs[2]; 707 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 708 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]], 709 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]* 710 // arrs_lhs = (S<float>*)lhs[2]; 711 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 712 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]], 713 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]* 714 715 // arrs_size = (size_t)lhs[3]; 716 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 717 // CHECK: [[ARRS_SIZE_VOID:%.+]] = load i8*, i8** [[ARRS_SIZE_REF]], 718 // CHECK: [[ARRS_SIZE:%.+]] = ptrtoint i8* [[ARRS_SIZE_VOID]] to i64 719 720 // arr_lhs[:] += arr_rhs[:]; 721 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]] 722 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]] 723 // CHECK: br i1 [[ISEMPTY]], 724 // CHECK: phi i32* 725 // CHECK: [[ADD:%.+]] = add nsw i32 % 726 // CHECK: store i32 [[ADD]], i32* % 727 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 728 // CHECK: br i1 [[DONE]], 729 730 // arrs_lhs = arrs_lhs.operator &(arrs_rhs); 731 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 732 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 733 // CHECK: br i1 [[ISEMPTY]], 734 // CHECK: phi [[S_FLOAT_TY]]* 735 // CHECK: [[AND:%.+]] = call nonnull align 4 dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* nonnull align 4 dereferenceable(4) %{{.+}}) 736 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 737 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %{{.+}}, i8* align 4 [[BITCAST]], i64 4, i1 false) 738 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 739 // CHECK: br i1 [[DONE]], 740 741 // CHECK: ret void 742 743 // 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]]]]* nonnull align 4 dereferenceable(160) %{{.+}}) 744 745 // CHECK: [[ARRS_PRIV:%.+]] = alloca [10 x [4 x [[S_FLOAT_TY]]]], 746 747 // Reduction list for runtime. 748 // CHECK: [[RED_LIST:%.+]] = alloca [3 x i8*], 749 750 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 751 752 // CHECK: [[ARR_SIZE:%.+]] = udiv exact i64 753 // CHECK: call i8* @llvm.stacksave() 754 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]], 755 756 // Check initialization of private copy. 757 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]] 758 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]] 759 // CHECK: br i1 [[ISEMPTY]], 760 // CHECK: phi i32* 761 // CHECK: store i32 0, i32* % 762 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 763 // CHECK: br i1 [[DONE]], 764 765 // Check initialization of private copy. 766 // 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 0 767 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40 768 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[BEGIN]], [[END]] 769 // CHECK: br i1 [[ISEMPTY]], 770 // CHECK: phi [[S_FLOAT_TY]]* 771 // CHECK: call void @_ZN1SIfEC1Ev([[S_FLOAT_TY]]* % 772 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 773 // CHECK: br i1 [[DONE]], 774 // CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]* 775 // CHECK: [[ARRS_PRIV_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]] to [[S_FLOAT_TY]]* 776 777 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 778 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 779 // CHECK: call void @__kmpc_for_static_init_4( 780 // Skip checks for internal operations. 781 // CHECK: call void @__kmpc_for_static_fini( 782 783 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 784 785 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 0 786 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8* 787 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]], 788 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 1 789 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8* 790 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]], 791 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 2 792 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV_BEGIN]] to i8* 793 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]], 794 795 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 796 797 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 798 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 799 // CHECK: [[BITCAST:%.+]] = bitcast [3 x i8*]* [[RED_LIST]] to i8* 800 // 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]]) 801 802 // switch(res) 803 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 804 // CHECK: i32 1, label %[[CASE1:.+]] 805 // CHECK: i32 2, label %[[CASE2:.+]] 806 // CHECK: ] 807 808 // case 1: 809 // CHECK: [[CASE1]] 810 811 // arr[:] += arr_reduction[:]; 812 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0:%.+]], i64 [[ARR_SIZE]] 813 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 814 // CHECK: br i1 [[ISEMPTY]], 815 // CHECK: phi i32* 816 // CHECK: [[ADD:%.+]] = add nsw i32 % 817 // CHECK: store i32 [[ADD]], i32* % 818 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 819 // CHECK: br i1 [[DONE]], 820 821 // arrs[:] = var.operator &(arrs_reduction[:]); 822 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40 823 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]] 824 // CHECK: br i1 [[ISEMPTY]], 825 // CHECK: phi [[S_FLOAT_TY]]* 826 // CHECK: [[AND:%.+]] = call nonnull align 4 dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* nonnull align 4 dereferenceable(4) %{{.+}}) 827 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 828 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %{{.+}}, i8* align 4 [[BITCAST]], i64 4, i1 false) 829 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 830 // CHECK: br i1 [[DONE]], 831 832 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 833 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 834 835 // break; 836 // CHECK: br label %[[RED_DONE]] 837 838 // case 2: 839 // CHECK: [[CASE2]] 840 841 // arr[:] += arr_reduction[:]; 842 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 843 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 844 // CHECK: br i1 [[ISEMPTY]], 845 // CHECK: phi i32* 846 // CHECK: atomicrmw add i32* %{{.+}}, i32 %{{.+}} monotonic 847 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 848 // CHECK: br i1 [[DONE]], 849 850 // arrs[:] = var.operator &(arrs_reduction[:]); 851 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40 852 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]] 853 // CHECK: br i1 [[ISEMPTY]], 854 // CHECK: phi [[S_FLOAT_TY]]* 855 // CHECK: call void @__kmpc_critical( 856 // CHECK: [[AND:%.+]] = call nonnull align 4 dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* nonnull align 4 dereferenceable(4) %{{.+}}) 857 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 858 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %{{.+}}, i8* align 4 [[BITCAST]], i64 4, i1 false) 859 // CHECK: call void @__kmpc_end_critical( 860 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 861 // CHECK: br i1 [[DONE]], 862 863 // break; 864 // CHECK: br label %[[RED_DONE]] 865 // CHECK: [[RED_DONE]] 866 867 // Check destruction of private copy. 868 // 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 0 869 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40 870 // CHECK: br 871 // CHECK: phi [[S_FLOAT_TY]]* 872 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* % 873 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[BEGIN]] 874 // CHECK: br i1 [[DONE]], 875 // CHECK: call void @llvm.stackrestore(i8* 876 // CHECK: call void @__kmpc_barrier( 877 878 // CHECK: ret void 879 880 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 881 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 882 // ... 883 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 884 // *(Type<n>-1*)rhs[<n>-1]); 885 // } 886 // CHECK: define internal void [[REDUCTION_FUNC]](i8* %0, i8* %1) 887 // arr_rhs = (int*)rhs[0]; 888 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 889 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]], 890 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32* 891 // arr_lhs = (int*)lhs[0]; 892 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 893 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]], 894 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32* 895 896 // arr_size = (size_t)lhs[1]; 897 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 898 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]], 899 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64 900 901 // arrs_rhs = (S<float>*)rhs[2]; 902 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 903 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]], 904 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]* 905 // arrs_lhs = (S<float>*)lhs[2]; 906 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 907 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]], 908 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]* 909 910 // arr_lhs[:] += arr_rhs[:]; 911 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]] 912 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]] 913 // CHECK: br i1 [[ISEMPTY]], 914 // CHECK: phi i32* 915 // CHECK: [[ADD:%.+]] = add nsw i32 % 916 // CHECK: store i32 [[ADD]], i32* % 917 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 918 // CHECK: br i1 [[DONE]], 919 920 // arrs_lhs = arrs_lhs.operator &(arrs_rhs); 921 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 40 922 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 923 // CHECK: br i1 [[ISEMPTY]], 924 // CHECK: phi [[S_FLOAT_TY]]* 925 // CHECK: [[AND:%.+]] = call nonnull align 4 dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* nonnull align 4 dereferenceable(4) %{{.+}}) 926 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 927 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %{{.+}}, i8* align 4 [[BITCAST]], i64 4, i1 false) 928 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 929 // CHECK: br i1 [[DONE]], 930 931 // CHECK: ret void 932 933 // CHECK: define internal void [[MAIN_MICROTASK3]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i32* {{.*}} %{{.+}}) 934 935 // CHECK: [[VLA1_ORIG_ADDR:%.+]] = alloca i64 936 // CHECK: [[VLA2_ORIG_ADDR:%.+]] = alloca i64 937 // CHECK: [[ARR_ORIG_ADDR:%.+]] = alloca i32*, 938 // CHECK: [[ARR_PRIV:%.+]] = alloca [1 x [2 x i32]], 939 940 // Reduction list for runtime. 941 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 942 943 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 944 // CHECK: [[VLA1:%.+]] = load i64, i64* [[VLA1_ORIG_ADDR]] 945 // CHECK: [[VLA2:%.+]] = load i64, i64* [[VLA2_ORIG_ADDR]] 946 // CHECK: [[ARR_ORIG:%.+]] = load i32*, i32** [[ARR_ORIG_ADDR]], 947 948 // CHECK: [[LOW_OFFSET:%.+]] = mul nsw i64 1, [[VLA2]] 949 // CHECK: [[ARRIDX:%.+]] = getelementptr inbounds i32, i32* [[ARR_ORIG]], i64 [[LOW_OFFSET]] 950 // CHECK: [[LOW:%.+]] = getelementptr inbounds i32, i32* [[ARRIDX]], i64 0 951 952 // CHECK: [[START:%.+]] = ptrtoint i32* [[ARR_ORIG]] to i64 953 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint i32* [[LOW]] to i64 954 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 955 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64) 956 // CHECK: [[ARR_PRIV_PTR:%.+]] = bitcast [1 x [2 x i32]]* [[ARR_PRIV]] to i32* 957 // CHECK: [[ARR_PRIV:%.+]] = getelementptr i32, i32* [[ARR_PRIV_PTR]], i64 [[OFFSET]] 958 959 // CHECK: ret void 960 961 // CHECK: define internal void [[MAIN_MICROTASK4]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** nonnull align 8 dereferenceable(8) %{{.+}}) 962 963 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***, 964 965 // Reduction list for runtime. 966 // CHECK: [[RED_LIST:%.+]] = alloca [2 x i8*], 967 968 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 969 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]], 970 971 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64 972 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW:%.+]] to i64 973 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]] 974 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 975 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1 976 // CHECK: call i8* @llvm.stacksave() 977 // CHECK: [[VAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]], 978 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 979 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], 980 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 981 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 982 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 983 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 984 // CHECK: [[PSEUDO_VAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR2_PRIV]], i64 [[OFFSET]] 985 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** % 986 // CHECK: store [[S_FLOAT_TY]]* [[PSEUDO_VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]] 987 // CHECK: ret void 988 989 // CHECK: define internal void [[MAIN_MICROTASK5]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** nonnull align 8 dereferenceable(8) %{{.+}}) 990 991 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***, 992 // CHECK: [[VAR2_PRIV:%.+]] = alloca [1 x [6 x [[S_FLOAT_TY]]]], 993 994 // Reduction list for runtime. 995 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 996 997 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 998 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]], 999 1000 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1001 // CHECK: [[ARRIDX:%.+]] = getelementptr inbounds [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], i64 1 1002 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[ARRIDX]], 1003 // CHECK: [[LOW:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LD]], i64 1 1004 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1005 1006 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1007 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], 1008 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1009 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1010 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1011 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1012 // CHECK: [[VAR2_PRIV_PTR:%.+]] = bitcast [1 x [6 x [[S_FLOAT_TY]]]]* [[VAR2_PRIV]] to [[S_FLOAT_TY]]* 1013 // CHECK: [[VAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR2_PRIV_PTR]], i64 [[OFFSET]] 1014 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** % 1015 // CHECK: store [[S_FLOAT_TY]]* [[VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]] 1016 // CHECK: ret void 1017 1018 // CHECK: define internal void [[MAIN_MICROTASK6]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** nonnull align 8 dereferenceable(8) %{{.+}}) 1019 1020 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***, 1021 // CHECK: [[VAR2_PRIV:%.+]] = alloca [1 x [6 x [[S_FLOAT_TY]]]], 1022 1023 // Reduction list for runtime. 1024 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1025 1026 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1027 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]], 1028 1029 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1030 // CHECK: [[ARRIDX:%.+]] = getelementptr inbounds [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], i64 1 1031 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[ARRIDX]], 1032 // CHECK: [[LOW:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LD]], i64 1 1033 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1034 1035 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1036 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], 1037 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1038 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1039 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1040 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1041 // CHECK: [[VAR2_PRIV_PTR:%.+]] = bitcast [1 x [6 x [[S_FLOAT_TY]]]]* [[VAR2_PRIV]] to [[S_FLOAT_TY]]* 1042 // CHECK: [[VAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR2_PRIV_PTR]], i64 [[OFFSET]] 1043 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** % 1044 // CHECK: store [[S_FLOAT_TY]]* [[VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]] 1045 // CHECK: ret void 1046 1047 // CHECK: define internal void [[MAIN_MICROTASK7]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** nonnull align 8 dereferenceable(8) %{{.+}}) 1048 1049 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***, 1050 // CHECK: [[VAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 1051 1052 // Reduction list for runtime. 1053 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1054 1055 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1056 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]], 1057 1058 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1059 // CHECK: [[ARRIDX:%.+]] = getelementptr inbounds [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], i64 1 1060 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[ARRIDX]], 1061 // CHECK: [[LOW:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LD]], i64 1 1062 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1063 1064 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1065 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], 1066 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1067 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1068 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1069 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1070 // CHECK: [[PSEUDO_VAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR2_PRIV]], i64 [[OFFSET]] 1071 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** % 1072 // CHECK: store [[S_FLOAT_TY]]* [[PSEUDO_VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]] 1073 // CHECK: ret void 1074 1075 // CHECK: define internal void [[MAIN_MICROTASK8]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [5 x [[S_FLOAT_TY]]]* nonnull align 4 dereferenceable(20) %{{.+}}) 1076 1077 // CHECK: [[VVAR2_ORIG_ADDR:%.+]] = alloca [5 x [[S_FLOAT_TY]]]*, 1078 // CHECK: [[VVAR2_PRIV:%.+]] = alloca [5 x [[S_FLOAT_TY]]], 1079 1080 // Reduction list for runtime. 1081 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1082 1083 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1084 // CHECK: [[VVAR2_ORIG:%.+]] = load [5 x [[S_FLOAT_TY]]]*, [5 x [[S_FLOAT_TY]]]** [[VVAR2_ORIG_ADDR]], 1085 1086 // CHECK: [[LOW:%.+]] = getelementptr inbounds [5 x [[S_FLOAT_TY]]], [5 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]], i64 0, i64 0 1087 // CHECK: [[ORIG_START:%.+]] = bitcast [5 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]] to [[S_FLOAT_TY]]* 1088 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1089 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1090 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1091 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1092 // CHECK: [[VVAR2_PRIV_PTR:%.+]] = bitcast [5 x [[S_FLOAT_TY]]]* [[VVAR2_PRIV]] to [[S_FLOAT_TY]]* 1093 // CHECK: [[VVAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VVAR2_PRIV_PTR]], i64 [[OFFSET]] 1094 // CHECK: ret void 1095 1096 // CHECK: define internal void [[MAIN_MICROTASK9]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* nonnull align 4 dereferenceable(16) %{{.+}}) 1097 1098 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 1099 // CHECK: [[VAR3_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]], 1100 1101 // Reduction list for runtime. 1102 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1103 1104 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1105 1106 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1107 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1108 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1109 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1110 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1111 1112 // CHECK: [[LOW:%.+]] = getelementptr inbounds [4 x [[S_FLOAT_TY]]], [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], i64 0, i64 1 1113 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1114 1115 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1116 // CHECK: [[ORIG_START:%.+]] = bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 1117 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1118 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1119 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1120 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1121 // CHECK: [[VAR3_PRIV_PTR:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]] to [[S_FLOAT_TY]]* 1122 // CHECK: [[PSEUDO_VAR3_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR3_PRIV_PTR]], i64 [[OFFSET]] 1123 // CHECK: [[VAR3_PRIV:%.+]] = bitcast [[S_FLOAT_TY]]* [[PSEUDO_VAR3_PRIV]] to [4 x [[S_FLOAT_TY]]]* 1124 1125 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 1126 1127 // CHECK: ret void 1128 1129 // CHECK: define internal void [[MAIN_MICROTASK10]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* nonnull align 4 dereferenceable(16) %{{.+}}) 1130 1131 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 1132 // CHECK: [[VAR3_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]], 1133 1134 // Reduction list for runtime. 1135 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1136 1137 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1138 1139 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1140 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1141 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1142 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1143 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1144 1145 // CHECK: [[LOW:%.+]] = getelementptr inbounds [4 x [[S_FLOAT_TY]]], [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], i64 0, i64 0 1146 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1147 1148 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1149 // CHECK: [[ORIG_START:%.+]] = bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 1150 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1151 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1152 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1153 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1154 // CHECK: [[VAR3_PRIV_PTR:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]] to [[S_FLOAT_TY]]* 1155 // CHECK: [[PSEUDO_VAR3_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR3_PRIV_PTR]], i64 [[OFFSET]] 1156 // CHECK: [[VAR3_PRIV:%.+]] = bitcast [[S_FLOAT_TY]]* [[PSEUDO_VAR3_PRIV]] to [4 x [[S_FLOAT_TY]]]* 1157 1158 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 1159 1160 // CHECK: ret void 1161 1162 // CHECK: define internal void [[MAIN_MICROTASK11]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* nonnull align 4 dereferenceable(16) %{{.+}}) 1163 1164 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 1165 1166 // Reduction list for runtime. 1167 // CHECK: [[RED_LIST:%.+]] = alloca [2 x i8*], 1168 1169 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1170 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1171 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1172 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1173 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1174 1175 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64 1176 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW:%.+]] to i64 1177 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]] 1178 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1179 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1 1180 // CHECK: call i8* @llvm.stacksave() 1181 // CHECK: [[VAR3_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]], 1182 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1183 // CHECK: [[ORIG_START:%.+]] = bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 1184 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1185 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1186 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1187 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1188 // CHECK: [[PSEUDO_VAR3_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR3_PRIV]], i64 [[OFFSET]] 1189 // CHECK: [[VAR3_PRIV:%.+]] = bitcast [[S_FLOAT_TY]]* [[PSEUDO_VAR3_PRIV]] to [4 x [[S_FLOAT_TY]]]* 1190 1191 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 1192 1193 // CHECK: ret void 1194 1195 // CHECK: define internal void [[MAIN_MICROTASK12]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* nonnull align 4 dereferenceable(16) %{{.+}}) 1196 1197 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 1198 // CHECK: [[VAR3_PRIV:%.+]] = alloca [4 x [[S_FLOAT_TY]]], 1199 1200 // Reduction list for runtime. 1201 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1202 1203 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1204 1205 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1206 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1207 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1208 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1209 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1210 // CHECK: getelementptr inbounds [4 x [[S_FLOAT_TY]]], [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], i32 0, i32 0 1211 // CHECK: getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 4 1212 1213 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 1214 // CHECK: bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 1215 1216 // CHECK: ret void 1217 1218 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT_42]]() 1219 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]], 1220 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]]) 1221 // 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 1222 // Not interested in this one: 1223 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4, 1224 // 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 1225 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]* 1226 // CHECK: ret 1227 // 1228 // CHECK: define internal void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i32* nonnull align 4 dereferenceable(4) %{{.+}}, [[S_INT_TY]]* nonnull align 4 dereferenceable(4) %{{.+}}, [[S_INT_TY]]* nonnull align 4 dereferenceable(4) %{{.+}}, i32* nonnull align 4 dereferenceable(4) %{{.+}}, [2 x i32]* nonnull align 4 dereferenceable(8) %{{.+}}, [2 x [[S_INT_TY]]]* nonnull align 4 dereferenceable(8) %{{.+}}) 1229 // CHECK: alloca i{{[0-9]+}}, 1230 // CHECK: alloca i{{[0-9]+}}, 1231 // CHECK: alloca i{{[0-9]+}}, 1232 // CHECK: alloca i{{[0-9]+}}, 1233 // CHECK: alloca i{{[0-9]+}}, 1234 // CHECK: alloca i{{[0-9]+}}, 1235 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 1236 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], 1237 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_INT_TY]], 1238 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca i{{[0-9]+}}, 1239 1240 // Reduction list for runtime. 1241 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 1242 1243 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1244 1245 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 1246 // CHECK: [[VAR_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 1247 // CHECK: [[VAR1_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 1248 // CHECK: [[T_VAR1_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 1249 1250 // For + reduction operation initial value of private variable is 0. 1251 // CHECK: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[T_VAR_PRIV]], 1252 1253 // For & reduction operation initial value of private variable is ones in all bits. 1254 // CHECK: [[VAR_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 1255 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR_PRIV]]) 1256 1257 // For && reduction operation initial value of private variable is 1.0. 1258 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR1_PRIV]]) 1259 1260 // For min reduction operation initial value of private variable is largest repesentable value. 1261 // CHECK: store i{{[0-9]+}} 2147483647, i{{[0-9]+}}* [[T_VAR1_PRIV]], 1262 1263 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 1264 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 1265 // CHECK: call void @__kmpc_for_static_init_4( 1266 // Skip checks for internal operations. 1267 // CHECK: call void @__kmpc_for_static_fini( 1268 1269 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 1270 1271 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 1272 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR_PRIV]] to i8* 1273 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 1274 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 1275 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR_PRIV]] to i8* 1276 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 1277 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 1278 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_PRIV]] to i8* 1279 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 1280 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 1281 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR1_PRIV]] to i8* 1282 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 1283 1284 // res = __kmpc_reduce_nowait(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 1285 1286 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 1287 // 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]]) 1288 1289 // switch(res) 1290 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 1291 // CHECK: i32 1, label %[[CASE1:.+]] 1292 // CHECK: i32 2, label %[[CASE2:.+]] 1293 // CHECK: ] 1294 1295 // case 1: 1296 // t_var += t_var_reduction; 1297 // CHECK: [[T_VAR_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_REF]], 1298 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]], 1299 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_VAL]], [[T_VAR_PRIV_VAL]] 1300 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_REF]], 1301 1302 // var = var.operator &(var_reduction); 1303 // CHECK: [[UP:%.+]] = call nonnull align 4 dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* nonnull align 4 dereferenceable(4) [[VAR_PRIV]]) 1304 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* 1305 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 1306 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 1307 1308 // var1 = var1.operator &&(var1_reduction); 1309 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) 1310 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1311 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 1312 // CHECK: [[TRUE]] 1313 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) 1314 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1315 // CHECK: br label %[[END2]] 1316 // CHECK: [[END2]] 1317 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 1318 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 1319 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 1320 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* 1321 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 1322 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 1323 1324 // t_var1 = min(t_var1, t_var1_reduction); 1325 // CHECK: [[T_VAR1_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_REF]], 1326 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]], 1327 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_VAL]], [[T_VAR1_PRIV_VAL]] 1328 // CHECK: br i1 [[CMP]] 1329 // CHECK: [[UP:%.+]] = phi i32 1330 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_REF]], 1331 1332 // __kmpc_end_reduce_nowait(<loc>, <gtid>, &<lock>); 1333 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 1334 1335 // break; 1336 // CHECK: br label %[[RED_DONE]] 1337 1338 // case 2: 1339 // t_var += t_var_reduction; 1340 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]] 1341 // CHECK: atomicrmw add i32* [[T_VAR_REF]], i32 [[T_VAR_PRIV_VAL]] monotonic 1342 1343 // var = var.operator &(var_reduction); 1344 // CHECK: call void @__kmpc_critical( 1345 // CHECK: [[UP:%.+]] = call nonnull align 4 dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* nonnull align 4 dereferenceable(4) [[VAR_PRIV]]) 1346 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* 1347 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 1348 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 1349 // CHECK: call void @__kmpc_end_critical( 1350 1351 // var1 = var1.operator &&(var1_reduction); 1352 // CHECK: call void @__kmpc_critical( 1353 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) 1354 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1355 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 1356 // CHECK: [[TRUE]] 1357 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) 1358 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1359 // CHECK: br label %[[END2]] 1360 // CHECK: [[END2]] 1361 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 1362 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 1363 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 1364 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* 1365 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 1366 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 1367 // CHECK: call void @__kmpc_end_critical( 1368 1369 // t_var1 = min(t_var1, t_var1_reduction); 1370 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]] 1371 // CHECK: atomicrmw min i32* [[T_VAR1_REF]], i32 [[T_VAR1_PRIV_VAL]] monotonic 1372 1373 // break; 1374 // CHECK: br label %[[RED_DONE]] 1375 // CHECK: [[RED_DONE]] 1376 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) 1377 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* 1378 // CHECK: ret void 1379 1380 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 1381 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 1382 // ... 1383 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 1384 // *(Type<n>-1*)rhs[<n>-1]); 1385 // } 1386 // CHECK: define internal void [[REDUCTION_FUNC]](i8* %0, i8* %1) 1387 // t_var_lhs = (i{{[0-9]+}}*)lhs[0]; 1388 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 1389 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 1390 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to i{{[0-9]+}}* 1391 // t_var_rhs = (i{{[0-9]+}}*)rhs[0]; 1392 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 1393 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 1394 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to i{{[0-9]+}}* 1395 1396 // var_lhs = (S<i{{[0-9]+}}>*)lhs[1]; 1397 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 1398 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 1399 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_INT_TY]]* 1400 // var_rhs = (S<i{{[0-9]+}}>*)rhs[1]; 1401 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 1402 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 1403 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_INT_TY]]* 1404 1405 // var1_lhs = (S<i{{[0-9]+}}>*)lhs[2]; 1406 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 1407 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 1408 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_INT_TY]]* 1409 // var1_rhs = (S<i{{[0-9]+}}>*)rhs[2]; 1410 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 1411 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 1412 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_INT_TY]]* 1413 1414 // t_var1_lhs = (i{{[0-9]+}}*)lhs[3]; 1415 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 1416 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 1417 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to i{{[0-9]+}}* 1418 // t_var1_rhs = (i{{[0-9]+}}*)rhs[3]; 1419 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 1420 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 1421 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to i{{[0-9]+}}* 1422 1423 // t_var_lhs += t_var_rhs; 1424 // CHECK: [[T_VAR_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_LHS]], 1425 // CHECK: [[T_VAR_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_RHS]], 1426 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_LHS_VAL]], [[T_VAR_RHS_VAL]] 1427 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_LHS]], 1428 1429 // var_lhs = var_lhs.operator &(var_rhs); 1430 // CHECK: [[UP:%.+]] = call nonnull align 4 dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_LHS]], [[S_INT_TY]]* nonnull align 4 dereferenceable(4) [[VAR_RHS]]) 1431 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_LHS]] to i8* 1432 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 1433 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 1434 1435 // var1_lhs = var1_lhs.operator &&(var1_rhs); 1436 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_LHS]]) 1437 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1438 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 1439 // CHECK: [[TRUE]] 1440 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_RHS]]) 1441 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1442 // CHECK: br label %[[END2]] 1443 // CHECK: [[END2]] 1444 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 1445 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 1446 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 1447 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_LHS]] to i8* 1448 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 1449 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 1450 1451 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 1452 // CHECK: [[T_VAR1_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_LHS]], 1453 // CHECK: [[T_VAR1_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_RHS]], 1454 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_LHS_VAL]], [[T_VAR1_RHS_VAL]] 1455 // CHECK: br i1 [[CMP]] 1456 // CHECK: [[UP:%.+]] = phi i32 1457 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_LHS]], 1458 // CHECK: ret void 1459 1460 // CHECK: define internal void [[TMAIN_MICROTASK2]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [42 x [[S_INT_TY]]]* nonnull align 4 dereferenceable(168) %{{.*}}, [2 x i32]* nonnull align 4 dereferenceable(8) %{{.*}}, i32* nonnull align 4 dereferenceable(4) %{{.*}}, [2 x [[S_INT_TY]]]* nonnull align 4 dereferenceable(8) %{{.*}}, [[S_INT_TY]]* nonnull align 4 dereferenceable(4) %{{.*}}) 1461 1462 // CHECK: [[ARR_ORIG_ADDR:%.+]] = alloca [42 x [[S_INT_TY]]]*, 1463 // CHECK: [[ARR_PRIV:%.+]] = alloca [40 x [[S_INT_TY]]], 1464 1465 // Reduction list for runtime. 1466 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1467 1468 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1469 1470 // CHECK: [[ARR_ORIG:%.+]] = load [42 x [[S_INT_TY]]]*, [42 x [[S_INT_TY]]]** [[ARR_ORIG_ADDR]], 1471 // CHECK: [[LOW:%.+]] = getelementptr inbounds [42 x [[S_INT_TY]]], [42 x [[S_INT_TY]]]* [[ARR_ORIG]], i64 0, i64 1 1472 // CHECK: [[ORIG_START:%.+]] = bitcast [42 x [[S_INT_TY]]]* [[ARR_ORIG]] to [[S_INT_TY]]* 1473 // CHECK: [[START:%.+]] = ptrtoint [[S_INT_TY]]* [[ORIG_START]] to i64 1474 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_INT_TY]]* [[LOW]] to i64 1475 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1476 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64) 1477 // CHECK: [[ARR_PRIV_PTR:%.+]] = bitcast [40 x [[S_INT_TY]]]* [[ARR_PRIV]] to [[S_INT_TY]]* 1478 // CHECK: [[PSEUDO_ARR_PRIV:%.+]] = getelementptr [[S_INT_TY]], [[S_INT_TY]]* [[ARR_PRIV_PTR]], i64 [[OFFSET]] 1479 // CHECK: [[ARR_PRIV:%.+]] = bitcast [[S_INT_TY]]* [[PSEUDO_ARR_PRIV]] to [42 x [[S_INT_TY]]]* 1480 1481 // CHECK: ret void 1482 1483 #endif 1484 1485