1 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck %s 2 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s 3 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s 4 // RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DLAMBDA -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -check-prefix=LAMBDA %s 5 // RUN: %clang_cc1 -verify -fopenmp -x c++ -fblocks -DBLOCKS -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -check-prefix=BLOCKS %s 6 // expected-no-diagnostics 7 #ifndef HEADER 8 #define HEADER 9 10 volatile double g, g_orig; 11 volatile double &g1 = g_orig; 12 13 template <class T> 14 struct S { 15 T f; 16 S(T a) : f(a + g) {} 17 S() : f(g) {} 18 operator T() { return T(); } 19 S &operator&(const S &) { return *this; } 20 ~S() {} 21 }; 22 23 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { float } 24 // CHECK-DAG: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} } 25 // CHECK-DAG: [[ATOMIC_REDUCE_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 26 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8* 27 // CHECK-DAG: [[REDUCTION_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 28 // CHECK-DAG: [[REDUCTION_LOCK:@.+]] = common global [8 x i32] zeroinitializer 29 30 template <typename T, int length> 31 T tmain() { 32 T t; 33 S<T> test; 34 T t_var = T(), t_var1; 35 T vec[] = {1, 2}; 36 S<T> s_arr[] = {1, 2}; 37 S<T> &var = test; 38 S<T> var1; 39 S<T> arr[length]; 40 #pragma omp parallel 41 #pragma omp for reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) nowait 42 for (int i = 0; i < 2; ++i) { 43 vec[i] = t_var; 44 s_arr[i] = var; 45 } 46 #pragma omp parallel 47 #pragma omp for reduction(&& : t_var) 48 for (int i = 0; i < 2; ++i) { 49 vec[i] = t_var; 50 s_arr[i] = var; 51 } 52 #pragma omp parallel 53 #pragma omp for reduction(+ : arr[1:length-2]) 54 for (int i = 0; i < 2; ++i) { 55 vec[i] = t_var; 56 s_arr[i] = var; 57 } 58 return T(); 59 } 60 61 extern S<float> **foo(); 62 63 int main() { 64 #ifdef LAMBDA 65 // LAMBDA: [[G:@.+]] = global double 66 // LAMBDA-LABEL: @main 67 // LAMBDA: call void [[OUTER_LAMBDA:@.+]]( 68 [&]() { 69 // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]]( 70 // LAMBDA: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}) 71 #pragma omp parallel 72 #pragma omp for reduction(+:g, g1) 73 for (int i = 0; i < 2; ++i) { 74 // LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}) 75 // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca double, 76 77 // Reduction list for runtime. 78 // LAMBDA: [[RED_LIST:%.+]] = alloca [2 x i8*], 79 80 // LAMBDA: store double 0.0{{.+}}, double* [[G_PRIVATE_ADDR]] 81 // LAMBDA: call void @__kmpc_for_static_init_4( 82 g = 1; 83 g1 = 1; 84 // LAMBDA: store double 1.0{{.+}}, double* [[G_PRIVATE_ADDR]], 85 // LAMBDA: [[G_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 86 // LAMBDA: store double* [[G_PRIVATE_ADDR]], double** [[G_PRIVATE_ADDR_REF]] 87 // LAMBDA: call void [[INNER_LAMBDA:@.+]](%{{.+}}* [[ARG]]) 88 // LAMBDA: call void @__kmpc_for_static_fini( 89 90 // LAMBDA: [[G_PRIV_REF:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[RED_LIST]], i64 0, i64 0 91 // LAMBDA: [[BITCAST:%.+]] = bitcast double* [[G_PRIVATE_ADDR]] to i8* 92 // LAMBDA: store i8* [[BITCAST]], i8** [[G_PRIV_REF]], 93 // LAMBDA: call i32 @__kmpc_reduce( 94 // LAMBDA: switch i32 %{{.+}}, label %[[REDUCTION_DONE:.+]] [ 95 // LAMBDA: i32 1, label %[[CASE1:.+]] 96 // LAMBDA: i32 2, label %[[CASE2:.+]] 97 // LAMBDA: [[CASE1]] 98 // LAMBDA: [[G_VAL:%.+]] = load double, double* [[G]] 99 // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 100 // LAMBDA: [[ADD:%.+]] = fadd double [[G_VAL]], [[G_PRIV_VAL]] 101 // LAMBDA: store double [[ADD]], double* [[G]] 102 // LAMBDA: call void @__kmpc_end_reduce( 103 // LAMBDA: br label %[[REDUCTION_DONE]] 104 // LAMBDA: [[CASE2]] 105 // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 106 // LAMBDA: fadd double 107 // LAMBDA: cmpxchg i64* 108 // LAMBDA: call void @__kmpc_end_reduce( 109 // LAMBDA: br label %[[REDUCTION_DONE]] 110 // LAMBDA: [[REDUCTION_DONE]] 111 // LAMBDA: ret void 112 [&]() { 113 // LAMBDA: define {{.+}} void [[INNER_LAMBDA]](%{{.+}}* [[ARG_PTR:%.+]]) 114 // LAMBDA: store %{{.+}}* [[ARG_PTR]], %{{.+}}** [[ARG_PTR_REF:%.+]], 115 g = 2; 116 g1 = 2; 117 // LAMBDA: [[ARG_PTR:%.+]] = load %{{.+}}*, %{{.+}}** [[ARG_PTR_REF]] 118 // LAMBDA: [[G_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 119 // LAMBDA: [[G_REF:%.+]] = load double*, double** [[G_PTR_REF]] 120 // LAMBDA: store double 2.0{{.+}}, double* [[G_REF]] 121 }(); 122 } 123 }(); 124 return 0; 125 #elif defined(BLOCKS) 126 // BLOCKS: [[G:@.+]] = global double 127 // BLOCKS-LABEL: @main 128 // BLOCKS: call void {{%.+}}(i8 129 ^{ 130 // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8* 131 // BLOCKS: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}) 132 #pragma omp parallel 133 #pragma omp for reduction(-:g, g1) 134 for (int i = 0; i < 2; ++i) { 135 // BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}) 136 // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca double, 137 138 // Reduction list for runtime. 139 // BLOCKS: [[RED_LIST:%.+]] = alloca [2 x i8*], 140 141 // BLOCKS: store double 0.0{{.+}}, double* [[G_PRIVATE_ADDR]] 142 g = 1; 143 g1 = 1; 144 // BLOCKS: call void @__kmpc_for_static_init_4( 145 // BLOCKS: store double 1.0{{.+}}, double* [[G_PRIVATE_ADDR]], 146 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 147 // BLOCKS: double* [[G_PRIVATE_ADDR]] 148 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 149 // BLOCKS: call void {{%.+}}(i8 150 // BLOCKS: call void @__kmpc_for_static_fini( 151 152 // BLOCKS: [[G_PRIV_REF:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[RED_LIST]], i64 0, i64 0 153 // BLOCKS: [[BITCAST:%.+]] = bitcast double* [[G_PRIVATE_ADDR]] to i8* 154 // BLOCKS: store i8* [[BITCAST]], i8** [[G_PRIV_REF]], 155 // BLOCKS: call i32 @__kmpc_reduce( 156 // BLOCKS: switch i32 %{{.+}}, label %[[REDUCTION_DONE:.+]] [ 157 // BLOCKS: i32 1, label %[[CASE1:.+]] 158 // BLOCKS: i32 2, label %[[CASE2:.+]] 159 // BLOCKS: [[CASE1]] 160 // BLOCKS: [[G_VAL:%.+]] = load double, double* [[G]] 161 // BLOCKS: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 162 // BLOCKS: [[ADD:%.+]] = fadd double [[G_VAL]], [[G_PRIV_VAL]] 163 // BLOCKS: store double [[ADD]], double* [[G]] 164 // BLOCKS: call void @__kmpc_end_reduce( 165 // BLOCKS: br label %[[REDUCTION_DONE]] 166 // BLOCKS: [[CASE2]] 167 // BLOCKS: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] 168 // BLOCKS: fadd double 169 // BLOCKS: cmpxchg i64* 170 // BLOCKS: call void @__kmpc_end_reduce( 171 // BLOCKS: br label %[[REDUCTION_DONE]] 172 // BLOCKS: [[REDUCTION_DONE]] 173 // BLOCKS: ret void 174 ^{ 175 // BLOCKS: define {{.+}} void {{@.+}}(i8* 176 g = 2; 177 g1 = 2; 178 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 179 // BLOCKS: store double 2.0{{.+}}, double* 180 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 181 // BLOCKS: ret 182 }(); 183 } 184 }(); 185 return 0; 186 #else 187 S<float> test; 188 float t_var = 0, t_var1; 189 int vec[] = {1, 2}; 190 S<float> s_arr[] = {1, 2, 3, 4}; 191 S<float> &var = test; 192 S<float> var1, arrs[10][4]; 193 S<float> **var2 = foo(); 194 S<float> vvar2[5]; 195 S<float> (&var3)[4] = s_arr; 196 #pragma omp parallel 197 #pragma omp for reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) 198 for (int i = 0; i < 2; ++i) { 199 vec[i] = t_var; 200 s_arr[i] = var; 201 } 202 int arr[10][vec[1]]; 203 #pragma omp parallel for reduction(+:arr[1][:vec[1]]) reduction(&:arrs[1:vec[1]][1:2]) 204 for (int i = 0; i < 10; ++i) 205 ++arr[1][i]; 206 #pragma omp parallel 207 #pragma omp for reduction(+:arr) reduction(&:arrs) 208 for (int i = 0; i < 10; ++i) 209 ++arr[1][i]; 210 // arr is a VLA, but the array section has constant length so we can generate a constant sized array! 211 #pragma omp parallel 212 #pragma omp for reduction(+:arr[1][0:2]) 213 for (int i = 0; i < 10; ++i) 214 ++arr[1][i]; 215 #pragma omp parallel 216 #pragma omp for reduction(& : var2[0 : 5][1 : 6]) 217 for (int i = 0; i < 10; ++i) 218 ; 219 #pragma omp parallel 220 #pragma omp for reduction(& : var2[1][1 : 6]) 221 for (int i = 0; i < 10; ++i) 222 ; 223 #pragma omp parallel 224 #pragma omp for reduction(& : var2[1 : 1][1 : 6]) 225 for (int i = 0; i < 10; ++i) 226 ; 227 #pragma omp parallel 228 #pragma omp for reduction(& : var2[1 : 1][1]) 229 for (int i = 0; i < 10; ++i) 230 ; 231 #pragma omp parallel 232 #pragma omp for reduction(& : vvar2[0 : 5]) 233 for (int i = 0; i < 10; ++i) 234 ; 235 #pragma omp parallel 236 #pragma omp for reduction(& : var3[1 : 2]) 237 for (int i = 0; i < 10; ++i) 238 ; 239 #pragma omp parallel 240 #pragma omp for reduction(& : var3[ : 2]) 241 for (int i = 0; i < 10; ++i) 242 ; 243 // TODO: The compiler should also be able to generate a constant sized array in this case! 244 #pragma omp parallel 245 #pragma omp for reduction(& : var3[2 : ]) 246 for (int i = 0; i < 10; ++i) 247 ; 248 #pragma omp parallel 249 #pragma omp for reduction(& : var3) 250 for (int i = 0; i < 10; ++i) 251 ; 252 return tmain<int, 42>(); 253 #endif 254 } 255 256 // CHECK: define {{.*}}i{{[0-9]+}} @main() 257 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]], 258 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]]) 259 // 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 260 // 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 261 // 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 262 // 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 263 // 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 264 // 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 265 // 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 266 // 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 267 // 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 268 // 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 269 // 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 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]+}}*, [4 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK11:@.+]] 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]+}}*, [4 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK12:@.+]] to void 272 // CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT_42:@.+]]() 273 // CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]* 274 // CHECK: ret 275 // 276 // CHECK: define internal void [[MAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, float* dereferenceable(4) %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}, float* dereferenceable(4) %{{.+}}, [2 x i32]* dereferenceable(8) %vec, [4 x [[S_FLOAT_TY]]]* dereferenceable(16) %{{.+}}) 277 // CHECK: [[T_VAR_PRIV:%.+]] = alloca float, 278 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 279 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 280 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca float, 281 282 // Reduction list for runtime. 283 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 284 285 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 286 287 // CHECK: [[T_VAR_REF:%.+]] = load float*, float** % 288 // CHECK: [[VAR1_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** % 289 // CHECK: [[T_VAR1_REF:%.+]] = load float*, float** % 290 291 // For + reduction operation initial value of private variable is 0. 292 // CHECK: store float 0.0{{.+}}, float* [[T_VAR_PRIV]], 293 294 // For & reduction operation initial value of private variable is ones in all bits. 295 // CHECK: [[VAR_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** % 296 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 297 298 // For && reduction operation initial value of private variable is 1.0. 299 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[VAR1_PRIV]]) 300 301 // For min reduction operation initial value of private variable is largest repesentable value. 302 // CHECK: store float 0x47EFFFFFE0000000, float* [[T_VAR1_PRIV]], 303 304 305 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 306 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 307 // CHECK: call void @__kmpc_for_static_init_4( 308 // Skip checks for internal operations. 309 // CHECK: call void @__kmpc_for_static_fini( 310 311 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 312 313 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 314 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR_PRIV]] to i8* 315 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 316 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 317 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_PRIV]] to i8* 318 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 319 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 320 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_PRIV]] to i8* 321 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 322 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 323 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR1_PRIV]] to i8* 324 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 325 326 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 327 328 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 329 // 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]]) 330 331 // switch(res) 332 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 333 // CHECK: i32 1, label %[[CASE1:.+]] 334 // CHECK: i32 2, label %[[CASE2:.+]] 335 // CHECK: ] 336 337 // case 1: 338 // t_var += t_var_reduction; 339 // CHECK: [[T_VAR_VAL:%.+]] = load float, float* [[T_VAR_REF]], 340 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load float, float* [[T_VAR_PRIV]], 341 // CHECK: [[UP:%.+]] = fadd float [[T_VAR_VAL]], [[T_VAR_PRIV_VAL]] 342 // CHECK: store float [[UP]], float* [[T_VAR_REF]], 343 344 // var = var.operator &(var_reduction); 345 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @{{.+}}([[S_FLOAT_TY]]* [[VAR_REF]], [[S_FLOAT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 346 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_REF]] to i8* 347 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[UP]] to i8* 348 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 349 350 // var1 = var1.operator &&(var1_reduction); 351 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_REF]]) 352 // CHECK: [[VAR1_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 353 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 354 // CHECK: [[TRUE]] 355 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_PRIV]]) 356 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 357 // CHECK: br label %[[END2]] 358 // CHECK: [[END2]] 359 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 360 // CHECK: [[CONV:%.+]] = uitofp i1 [[COND_LVALUE]] to float 361 // CHECK: call void @{{.+}}([[S_FLOAT_TY]]* [[COND_LVALUE:%.+]], float [[CONV]]) 362 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_REF]] to i8* 363 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[COND_LVALUE]] to i8* 364 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 365 366 // t_var1 = min(t_var1, t_var1_reduction); 367 // CHECK: [[T_VAR1_VAL:%.+]] = load float, float* [[T_VAR1_REF]], 368 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load float, float* [[T_VAR1_PRIV]], 369 // CHECK: [[CMP:%.+]] = fcmp olt float [[T_VAR1_VAL]], [[T_VAR1_PRIV_VAL]] 370 // CHECK: br i1 [[CMP]] 371 // CHECK: [[UP:%.+]] = phi float 372 // CHECK: store float [[UP]], float* [[T_VAR1_REF]], 373 374 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 375 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 376 377 // break; 378 // CHECK: br label %[[RED_DONE]] 379 380 // case 2: 381 // t_var += t_var_reduction; 382 // CHECK: load float, float* [[T_VAR_PRIV]] 383 // CHECK: [[T_VAR_REF_INT:%.+]] = bitcast float* [[T_VAR_REF]] to i32* 384 // CHECK: [[OLD1:%.+]] = load atomic i32, i32* [[T_VAR_REF_INT]] monotonic, 385 // CHECK: br label %[[CONT:.+]] 386 // CHECK: [[CONT]] 387 // CHECK: [[ORIG_OLD_INT:%.+]] = phi i32 [ [[OLD1]], %{{.+}} ], [ [[OLD2:%.+]], %[[CONT]] ] 388 // CHECK: fadd float 389 // CHECK: [[UP_INT:%.+]] = load i32, i32* 390 // CHECK: [[T_VAR_REF_INT:%.+]] = bitcast float* [[T_VAR_REF]] to i32* 391 // CHECK: [[RES:%.+]] = cmpxchg i32* [[T_VAR_REF_INT]], i32 [[ORIG_OLD_INT]], i32 [[UP_INT]] monotonic monotonic 392 // CHECK: [[OLD2:%.+]] = extractvalue { i32, i1 } [[RES]], 0 393 // CHECK: [[SUCCESS_FAIL:%.+]] = extractvalue { i32, i1 } [[RES]], 1 394 // CHECK: br i1 [[SUCCESS_FAIL]], label %[[ATOMIC_DONE:.+]], label %[[CONT]] 395 // CHECK: [[ATOMIC_DONE]] 396 397 // var = var.operator &(var_reduction); 398 // CHECK: call void @__kmpc_critical( 399 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @{{.+}}([[S_FLOAT_TY]]* [[VAR_REF]], [[S_FLOAT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 400 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_REF]] to i8* 401 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[UP]] to i8* 402 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 403 // CHECK: call void @__kmpc_end_critical( 404 405 // var1 = var1.operator &&(var1_reduction); 406 // CHECK: call void @__kmpc_critical( 407 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_REF]]) 408 // CHECK: [[VAR1_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 409 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 410 // CHECK: [[TRUE]] 411 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_PRIV]]) 412 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 413 // CHECK: br label %[[END2]] 414 // CHECK: [[END2]] 415 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 416 // CHECK: [[CONV:%.+]] = uitofp i1 [[COND_LVALUE]] to float 417 // CHECK: call void @{{.+}}([[S_FLOAT_TY]]* [[COND_LVALUE:%.+]], float [[CONV]]) 418 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_REF]] to i8* 419 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[COND_LVALUE]] to i8* 420 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 421 // CHECK: call void @__kmpc_end_critical( 422 423 // t_var1 = min(t_var1, t_var1_reduction); 424 // CHECK: load float, float* [[T_VAR1_PRIV]] 425 // CHECK: [[T_VAR1_REF_INT:%.+]] = bitcast float* [[T_VAR1_REF]] to i32* 426 // CHECK: [[OLD1:%.+]] = load atomic i32, i32* [[T_VAR1_REF_INT]] monotonic, 427 // CHECK: br label %[[CONT:.+]] 428 // CHECK: [[CONT]] 429 // CHECK: [[ORIG_OLD_INT:%.+]] = phi i32 [ [[OLD1]], %{{.+}} ], [ [[OLD2:%.+]], %{{.+}} ] 430 // CHECK: [[CMP:%.+]] = fcmp olt float 431 // CHECK: br i1 [[CMP]] 432 // CHECK: phi float 433 // CHECK: [[UP_INT:%.+]] = load i32 434 // CHECK: [[T_VAR1_REF_INT:%.+]] = bitcast float* [[T_VAR1_REF]] to i32* 435 // CHECK: [[RES:%.+]] = cmpxchg i32* [[T_VAR1_REF_INT]], i32 [[ORIG_OLD_INT]], i32 [[UP_INT]] monotonic monotonic 436 // CHECK: [[OLD2:%.+]] = extractvalue { i32, i1 } [[RES]], 0 437 // CHECK: [[SUCCESS_FAIL:%.+]] = extractvalue { i32, i1 } [[RES]], 1 438 // CHECK: br i1 [[SUCCESS_FAIL]], label %[[ATOMIC_DONE:.+]], label %[[CONT]] 439 // CHECK: [[ATOMIC_DONE]] 440 441 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 442 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 443 444 // break; 445 // CHECK: br label %[[RED_DONE]] 446 // CHECK: [[RED_DONE]] 447 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 448 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* 449 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 450 451 // CHECK: ret void 452 453 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 454 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 455 // ... 456 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 457 // *(Type<n>-1*)rhs[<n>-1]); 458 // } 459 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 460 // t_var_lhs = (float*)lhs[0]; 461 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 462 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 463 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to float* 464 // t_var_rhs = (float*)rhs[0]; 465 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 466 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 467 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to float* 468 469 // var_lhs = (S<float>*)lhs[1]; 470 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 471 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 472 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_FLOAT_TY]]* 473 // var_rhs = (S<float>*)rhs[1]; 474 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 475 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 476 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_FLOAT_TY]]* 477 478 // var1_lhs = (S<float>*)lhs[2]; 479 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 480 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 481 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_FLOAT_TY]]* 482 // var1_rhs = (S<float>*)rhs[2]; 483 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 484 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 485 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_FLOAT_TY]]* 486 487 // t_var1_lhs = (float*)lhs[3]; 488 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 489 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 490 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to float* 491 // t_var1_rhs = (float*)rhs[3]; 492 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 493 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 494 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to float* 495 496 // t_var_lhs += t_var_rhs; 497 // CHECK: [[T_VAR_LHS_VAL:%.+]] = load float, float* [[T_VAR_LHS]], 498 // CHECK: [[T_VAR_RHS_VAL:%.+]] = load float, float* [[T_VAR_RHS]], 499 // CHECK: [[UP:%.+]] = fadd float [[T_VAR_LHS_VAL]], [[T_VAR_RHS_VAL]] 500 // CHECK: store float [[UP]], float* [[T_VAR_LHS]], 501 502 // var_lhs = var_lhs.operator &(var_rhs); 503 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @{{.+}}([[S_FLOAT_TY]]* [[VAR_LHS]], [[S_FLOAT_TY]]* dereferenceable(4) [[VAR_RHS]]) 504 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_LHS]] to i8* 505 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[UP]] to i8* 506 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 507 508 // var1_lhs = var1_lhs.operator &&(var1_rhs); 509 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_LHS]]) 510 // CHECK: [[VAR1_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 511 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 512 // CHECK: [[TRUE]] 513 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_RHS]]) 514 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 515 // CHECK: br label %[[END2]] 516 // CHECK: [[END2]] 517 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 518 // CHECK: [[CONV:%.+]] = uitofp i1 [[COND_LVALUE]] to float 519 // CHECK: call void @{{.+}}([[S_FLOAT_TY]]* [[COND_LVALUE:%.+]], float [[CONV]]) 520 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_LHS]] to i8* 521 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[COND_LVALUE]] to i8* 522 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 523 524 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 525 // CHECK: [[T_VAR1_LHS_VAL:%.+]] = load float, float* [[T_VAR1_LHS]], 526 // CHECK: [[T_VAR1_RHS_VAL:%.+]] = load float, float* [[T_VAR1_RHS]], 527 // CHECK: [[CMP:%.+]] = fcmp olt float [[T_VAR1_LHS_VAL]], [[T_VAR1_RHS_VAL]] 528 // CHECK: br i1 [[CMP]] 529 // CHECK: [[UP:%.+]] = phi float 530 // CHECK: store float [[UP]], float* [[T_VAR1_LHS]], 531 // CHECK: ret void 532 533 // 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(160) %{{.+}}) 534 535 // Reduction list for runtime. 536 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 537 538 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 539 // CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}** 540 // CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}** [[ARR_ADDR:%.+]], 541 // CHECK: [[ARR:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[ARR_ADDR]], 542 543 // CHECK: [[UB_CAST:%.+]] = ptrtoint i32* [[UB1_UP:%.+]] to i64 544 // CHECK: [[LB_CAST:%.+]] = ptrtoint i32* [[LB1_0:%.+]] to i64 545 // CHECK: [[DIFF:%.+]] = sub i64 [[UB_CAST]], [[LB_CAST]] 546 // CHECK: [[SIZE_1:%.+]] = sdiv exact i64 [[DIFF]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64) 547 // CHECK: [[ARR_SIZE:%.+]] = add nuw i64 [[SIZE_1]], 1 548 // CHECK: call i8* @llvm.stacksave() 549 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]], 550 551 // Check initialization of private copy. 552 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]] 553 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]] 554 // CHECK: br i1 [[ISEMPTY]], 555 // CHECK: phi i32* 556 // CHECK: store i32 0, i32* % 557 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 558 // CHECK: br i1 [[DONE]], 559 560 // CHECK: [[ARRS_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[ARRS_SIZE:%.+]], 561 562 // Check initialization of private copy. 563 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]] 564 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]] 565 // CHECK: br i1 [[ISEMPTY]], 566 // CHECK: phi [[S_FLOAT_TY]]* 567 // CHECK: call void @_ZN1SIfEC1Ev([[S_FLOAT_TY]]* % 568 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 569 // CHECK: br i1 [[DONE]], 570 571 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 572 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 573 // CHECK: call void @__kmpc_for_static_init_4( 574 // Skip checks for internal operations. 575 // CHECK: call void @__kmpc_for_static_fini( 576 577 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 578 579 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 580 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8* 581 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]], 582 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 583 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8* 584 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]], 585 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 586 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV]] to i8* 587 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]], 588 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 589 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARRS_SIZE]] to i8* 590 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_SIZE_REF]], 591 592 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 593 594 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 595 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 596 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 597 // 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]]) 598 599 // switch(res) 600 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 601 // CHECK: i32 1, label %[[CASE1:.+]] 602 // CHECK: i32 2, label %[[CASE2:.+]] 603 // CHECK: ] 604 605 // case 1: 606 // CHECK: [[CASE1]] 607 608 // arr[:] += arr_reduction[:]; 609 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 610 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 611 // CHECK: br i1 [[ISEMPTY]], 612 // CHECK: phi i32* 613 // CHECK: [[ADD:%.+]] = add nsw i32 % 614 // CHECK: store i32 [[ADD]], i32* % 615 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 616 // CHECK: br i1 [[DONE]], 617 618 // arrs[:] = var.operator &(arrs_reduction[:]); 619 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 620 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 621 // CHECK: br i1 [[ISEMPTY]], 622 // CHECK: phi [[S_FLOAT_TY]]* 623 // CHECK: [[AND:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}) 624 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 625 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %{{.+}}, i8* [[BITCAST]], i64 4, i32 4, i1 false) 626 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 627 // CHECK: br i1 [[DONE]], 628 629 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 630 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 631 632 // break; 633 // CHECK: br label %[[RED_DONE]] 634 635 // case 2: 636 // CHECK: [[CASE2]] 637 638 // arr[:] += arr_reduction[:]; 639 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 640 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 641 // CHECK: br i1 [[ISEMPTY]], 642 // CHECK: phi i32* 643 // CHECK: atomicrmw add i32* %{{.+}}, i32 %{{.+}} monotonic 644 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 645 // CHECK: br i1 [[DONE]], 646 647 // arrs[:] = var.operator &(arrs_reduction[:]); 648 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 649 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 650 // CHECK: br i1 [[ISEMPTY]], 651 // CHECK: phi [[S_FLOAT_TY]]* 652 // CHECK: call void @__kmpc_critical( 653 // CHECK: [[AND:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}) 654 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 655 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %{{.+}}, i8* [[BITCAST]], i64 4, i32 4, i1 false) 656 // CHECK: call void @__kmpc_end_critical( 657 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 658 // CHECK: br i1 [[DONE]], 659 660 // break; 661 // CHECK: br label %[[RED_DONE]] 662 // CHECK: [[RED_DONE]] 663 664 // Check destruction of private copy. 665 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]] 666 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]] 667 // CHECK: br i1 [[ISEMPTY]], 668 // CHECK: phi [[S_FLOAT_TY]]* 669 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* % 670 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[ARRS_PRIV]] 671 // CHECK: br i1 [[DONE]], 672 // CHECK: call void @llvm.stackrestore(i8* 673 674 // CHECK: ret void 675 676 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 677 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 678 // ... 679 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 680 // *(Type<n>-1*)rhs[<n>-1]); 681 // } 682 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 683 // arr_rhs = (int*)rhs[0]; 684 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 685 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]], 686 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32* 687 // arr_lhs = (int*)lhs[0]; 688 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 689 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]], 690 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32* 691 692 // arr_size = (size_t)lhs[1]; 693 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 694 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]], 695 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64 696 697 // arrs_rhs = (S<float>*)rhs[2]; 698 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 699 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]], 700 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]* 701 // arrs_lhs = (S<float>*)lhs[2]; 702 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 703 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]], 704 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]* 705 706 // arrs_size = (size_t)lhs[3]; 707 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 708 // CHECK: [[ARRS_SIZE_VOID:%.+]] = load i8*, i8** [[ARRS_SIZE_REF]], 709 // CHECK: [[ARRS_SIZE:%.+]] = ptrtoint i8* [[ARRS_SIZE_VOID]] to i64 710 711 // arr_lhs[:] += arr_rhs[:]; 712 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]] 713 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]] 714 // CHECK: br i1 [[ISEMPTY]], 715 // CHECK: phi i32* 716 // CHECK: [[ADD:%.+]] = add nsw i32 % 717 // CHECK: store i32 [[ADD]], i32* % 718 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 719 // CHECK: br i1 [[DONE]], 720 721 // arrs_lhs = arrs_lhs.operator &(arrs_rhs); 722 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 723 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 724 // CHECK: br i1 [[ISEMPTY]], 725 // CHECK: phi [[S_FLOAT_TY]]* 726 // CHECK: [[AND:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}) 727 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 728 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %{{.+}}, i8* [[BITCAST]], i64 4, i32 4, i1 false) 729 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 730 // CHECK: br i1 [[DONE]], 731 732 // CHECK: ret void 733 734 // 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(160) %{{.+}}) 735 736 // CHECK: [[ARRS_PRIV:%.+]] = alloca [10 x [4 x [[S_FLOAT_TY]]]], 737 738 // Reduction list for runtime. 739 // CHECK: [[RED_LIST:%.+]] = alloca [3 x i8*], 740 741 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 742 743 // CHECK: [[ARR_SIZE:%.+]] = udiv exact i64 744 // CHECK: call i8* @llvm.stacksave() 745 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]], 746 747 // Check initialization of private copy. 748 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]] 749 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]] 750 // CHECK: br i1 [[ISEMPTY]], 751 // CHECK: phi i32* 752 // CHECK: store i32 0, i32* % 753 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 754 // CHECK: br i1 [[DONE]], 755 756 // Check initialization of private copy. 757 // 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 0 758 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40 759 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[BEGIN]], [[END]] 760 // CHECK: br i1 [[ISEMPTY]], 761 // CHECK: phi [[S_FLOAT_TY]]* 762 // CHECK: call void @_ZN1SIfEC1Ev([[S_FLOAT_TY]]* % 763 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 764 // CHECK: br i1 [[DONE]], 765 // CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]* 766 // CHECK: [[ARRS_PRIV_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]] to [[S_FLOAT_TY]]* 767 768 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 769 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 770 // CHECK: call void @__kmpc_for_static_init_4( 771 // Skip checks for internal operations. 772 // CHECK: call void @__kmpc_for_static_fini( 773 774 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 775 776 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 0 777 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8* 778 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]], 779 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 1 780 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8* 781 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]], 782 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 2 783 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV_BEGIN]] to i8* 784 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]], 785 786 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 787 788 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 789 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 790 // CHECK: [[BITCAST:%.+]] = bitcast [3 x i8*]* [[RED_LIST]] to i8* 791 // 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]]) 792 793 // switch(res) 794 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 795 // CHECK: i32 1, label %[[CASE1:.+]] 796 // CHECK: i32 2, label %[[CASE2:.+]] 797 // CHECK: ] 798 799 // case 1: 800 // CHECK: [[CASE1]] 801 802 // arr[:] += arr_reduction[:]; 803 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0:%.+]], i64 [[ARR_SIZE]] 804 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 805 // CHECK: br i1 [[ISEMPTY]], 806 // CHECK: phi i32* 807 // CHECK: [[ADD:%.+]] = add nsw i32 % 808 // CHECK: store i32 [[ADD]], i32* % 809 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 810 // CHECK: br i1 [[DONE]], 811 812 // arrs[:] = var.operator &(arrs_reduction[:]); 813 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40 814 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]] 815 // CHECK: br i1 [[ISEMPTY]], 816 // CHECK: phi [[S_FLOAT_TY]]* 817 // CHECK: [[AND:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}) 818 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 819 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %{{.+}}, i8* [[BITCAST]], i64 4, i32 4, i1 false) 820 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 821 // CHECK: br i1 [[DONE]], 822 823 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 824 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 825 826 // break; 827 // CHECK: br label %[[RED_DONE]] 828 829 // case 2: 830 // CHECK: [[CASE2]] 831 832 // arr[:] += arr_reduction[:]; 833 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 834 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 835 // CHECK: br i1 [[ISEMPTY]], 836 // CHECK: phi i32* 837 // CHECK: atomicrmw add i32* %{{.+}}, i32 %{{.+}} monotonic 838 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 839 // CHECK: br i1 [[DONE]], 840 841 // arrs[:] = var.operator &(arrs_reduction[:]); 842 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40 843 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]] 844 // CHECK: br i1 [[ISEMPTY]], 845 // CHECK: phi [[S_FLOAT_TY]]* 846 // CHECK: call void @__kmpc_critical( 847 // CHECK: [[AND:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}) 848 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 849 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %{{.+}}, i8* [[BITCAST]], i64 4, i32 4, i1 false) 850 // CHECK: call void @__kmpc_end_critical( 851 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 852 // CHECK: br i1 [[DONE]], 853 854 // break; 855 // CHECK: br label %[[RED_DONE]] 856 // CHECK: [[RED_DONE]] 857 858 // Check destruction of private copy. 859 // 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 0 860 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40 861 // CHECK: br 862 // CHECK: phi [[S_FLOAT_TY]]* 863 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* % 864 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[BEGIN]] 865 // CHECK: br i1 [[DONE]], 866 // CHECK: call void @llvm.stackrestore(i8* 867 // CHECK: call void @__kmpc_barrier( 868 869 // CHECK: ret void 870 871 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 872 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 873 // ... 874 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 875 // *(Type<n>-1*)rhs[<n>-1]); 876 // } 877 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 878 // arr_rhs = (int*)rhs[0]; 879 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 880 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]], 881 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32* 882 // arr_lhs = (int*)lhs[0]; 883 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 884 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]], 885 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32* 886 887 // arr_size = (size_t)lhs[1]; 888 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 889 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]], 890 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64 891 892 // arrs_rhs = (S<float>*)rhs[2]; 893 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 894 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]], 895 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]* 896 // arrs_lhs = (S<float>*)lhs[2]; 897 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 898 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]], 899 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]* 900 901 // arr_lhs[:] += arr_rhs[:]; 902 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]] 903 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]] 904 // CHECK: br i1 [[ISEMPTY]], 905 // CHECK: phi i32* 906 // CHECK: [[ADD:%.+]] = add nsw i32 % 907 // CHECK: store i32 [[ADD]], i32* % 908 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 909 // CHECK: br i1 [[DONE]], 910 911 // arrs_lhs = arrs_lhs.operator &(arrs_rhs); 912 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 40 913 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 914 // CHECK: br i1 [[ISEMPTY]], 915 // CHECK: phi [[S_FLOAT_TY]]* 916 // CHECK: [[AND:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}) 917 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 918 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %{{.+}}, i8* [[BITCAST]], i64 4, i32 4, i1 false) 919 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 920 // CHECK: br i1 [[DONE]], 921 922 // CHECK: ret void 923 924 // CHECK: define internal void [[MAIN_MICROTASK3]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i32* {{.*}} %{{.+}}) 925 926 // CHECK: [[VLA1_ORIG_ADDR:%.+]] = alloca i64 927 // CHECK: [[VLA2_ORIG_ADDR:%.+]] = alloca i64 928 // CHECK: [[ARR_ORIG_ADDR:%.+]] = alloca i32*, 929 // CHECK: [[ARR_PRIV:%.+]] = alloca [1 x [2 x i32]], 930 931 // Reduction list for runtime. 932 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 933 934 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 935 // CHECK: [[VLA1:%.+]] = load i64, i64* [[VLA1_ORIG_ADDR]] 936 // CHECK: [[VLA2:%.+]] = load i64, i64* [[VLA2_ORIG_ADDR]] 937 // CHECK: [[ARR_ORIG:%.+]] = load i32*, i32** [[ARR_ORIG_ADDR]], 938 939 // CHECK: [[LOW_OFFSET:%.+]] = mul nsw i64 1, [[VLA2]] 940 // CHECK: [[ARRIDX:%.+]] = getelementptr inbounds i32, i32* [[ARR_ORIG]], i64 [[LOW_OFFSET]] 941 // CHECK: [[LOW:%.+]] = getelementptr inbounds i32, i32* [[ARRIDX]], i64 0 942 943 // CHECK: [[START:%.+]] = ptrtoint i32* [[ARR_ORIG]] to i64 944 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint i32* [[LOW]] to i64 945 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 946 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64) 947 // CHECK: [[PSEUDO_ARR_PRIV:%.+]] = getelementptr [1 x [2 x i32]], [1 x [2 x i32]]* [[ARR_PRIV]], i64 [[OFFSET]] 948 // CHECK: [[ARR_PRIV:%.+]] = bitcast [1 x [2 x i32]]* [[PSEUDO_ARR_PRIV]] to i32* 949 950 // CHECK: ret void 951 952 // CHECK: define internal void [[MAIN_MICROTASK4]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** dereferenceable(8) %{{.+}}) 953 954 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***, 955 956 // Reduction list for runtime. 957 // CHECK: [[RED_LIST:%.+]] = alloca [2 x i8*], 958 959 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 960 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]], 961 962 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64 963 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW:%.+]] to i64 964 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]] 965 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 966 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1 967 // CHECK: call i8* @llvm.stacksave() 968 // CHECK: [[VAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]], 969 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 970 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], 971 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 972 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 973 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 974 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 975 // CHECK: [[PSEUDO_VAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR2_PRIV]], i64 [[OFFSET]] 976 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** % 977 // CHECK: store [[S_FLOAT_TY]]* [[PSEUDO_VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]] 978 // CHECK: ret void 979 980 // CHECK: define internal void [[MAIN_MICROTASK5]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** dereferenceable(8) %{{.+}}) 981 982 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***, 983 // CHECK: [[VAR2_PRIV:%.+]] = alloca [1 x [6 x [[S_FLOAT_TY]]]], 984 985 // Reduction list for runtime. 986 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 987 988 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 989 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]], 990 991 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 992 // CHECK: [[ARRIDX:%.+]] = getelementptr inbounds [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], i64 1 993 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[ARRIDX]], 994 // CHECK: [[LOW:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LD]], i64 1 995 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 996 997 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 998 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], 999 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1000 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1001 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1002 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1003 // CHECK: [[PSEUDO_VAR2_PRIV:%.+]] = getelementptr [1 x [6 x [[S_FLOAT_TY]]]], [1 x [6 x [[S_FLOAT_TY]]]]* [[VAR2_PRIV]], i64 [[OFFSET]] 1004 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** % 1005 // CHECK: [[VAR2_PRIV:%.+]] = bitcast [1 x [6 x [[S_FLOAT_TY]]]]* [[PSEUDO_VAR2_PRIV]] to [[S_FLOAT_TY]]* 1006 // CHECK: store [[S_FLOAT_TY]]* [[VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]] 1007 // CHECK: ret void 1008 1009 // CHECK: define internal void [[MAIN_MICROTASK6]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** dereferenceable(8) %{{.+}}) 1010 1011 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***, 1012 // CHECK: [[VAR2_PRIV:%.+]] = alloca [1 x [6 x [[S_FLOAT_TY]]]], 1013 1014 // Reduction list for runtime. 1015 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1016 1017 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1018 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]], 1019 1020 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1021 // CHECK: [[ARRIDX:%.+]] = getelementptr inbounds [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], i64 1 1022 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[ARRIDX]], 1023 // CHECK: [[LOW:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LD]], i64 1 1024 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1025 1026 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1027 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], 1028 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1029 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1030 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1031 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1032 // CHECK: [[PSEUDO_VAR2_PRIV:%.+]] = getelementptr [1 x [6 x [[S_FLOAT_TY]]]], [1 x [6 x [[S_FLOAT_TY]]]]* [[VAR2_PRIV]], i64 [[OFFSET]] 1033 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** % 1034 // CHECK: [[VAR2_PRIV:%.+]] = bitcast [1 x [6 x [[S_FLOAT_TY]]]]* [[PSEUDO_VAR2_PRIV]] to [[S_FLOAT_TY]]* 1035 // CHECK: store [[S_FLOAT_TY]]* [[VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]] 1036 // CHECK: ret void 1037 1038 // CHECK: define internal void [[MAIN_MICROTASK7]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** dereferenceable(8) %{{.+}}) 1039 1040 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***, 1041 // CHECK: [[VAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 1042 1043 // Reduction list for runtime. 1044 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1045 1046 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1047 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]], 1048 1049 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1050 // CHECK: [[ARRIDX:%.+]] = getelementptr inbounds [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], i64 1 1051 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[ARRIDX]], 1052 // CHECK: [[LOW:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LD]], i64 1 1053 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1054 1055 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1056 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], 1057 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1058 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1059 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1060 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1061 // CHECK: [[PSEUDO_VAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR2_PRIV]], i64 [[OFFSET]] 1062 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** % 1063 // CHECK: store [[S_FLOAT_TY]]* [[PSEUDO_VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]] 1064 // CHECK: ret void 1065 1066 // CHECK: define internal void [[MAIN_MICROTASK8]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [5 x [[S_FLOAT_TY]]]* dereferenceable(20) %{{.+}}) 1067 1068 // CHECK: [[VVAR2_ORIG_ADDR:%.+]] = alloca [5 x [[S_FLOAT_TY]]]*, 1069 // CHECK: [[VVAR2_PRIV:%.+]] = alloca [5 x [[S_FLOAT_TY]]], 1070 1071 // Reduction list for runtime. 1072 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1073 1074 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1075 // CHECK: [[VVAR2_ORIG:%.+]] = load [5 x [[S_FLOAT_TY]]]*, [5 x [[S_FLOAT_TY]]]** [[VVAR2_ORIG_ADDR]], 1076 1077 // CHECK: [[LOW:%.+]] = getelementptr inbounds [5 x [[S_FLOAT_TY]]], [5 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]], i64 0, i64 0 1078 // CHECK: [[ORIG_START:%.+]] = bitcast [5 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]] to [[S_FLOAT_TY]]* 1079 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1080 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1081 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1082 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1083 // CHECK: [[PSEUDO_VVAR2_PRIV:%.+]] = getelementptr [5 x [[S_FLOAT_TY]]], [5 x [[S_FLOAT_TY]]]* [[VVAR2_PRIV]], i64 [[OFFSET]] 1084 // CHECK: ret void 1085 1086 // CHECK: define internal void [[MAIN_MICROTASK9]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* dereferenceable(16) %{{.+}}) 1087 1088 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 1089 // CHECK: [[VAR3_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]], 1090 1091 // Reduction list for runtime. 1092 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1093 1094 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1095 1096 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1097 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1098 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1099 1100 // CHECK: [[LOW:%.+]] = getelementptr inbounds [4 x [[S_FLOAT_TY]]], [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], i64 0, i64 1 1101 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1102 1103 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1104 // CHECK: [[ORIG_START:%.+]] = bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 1105 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1106 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1107 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1108 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1109 // CHECK: [[PSEUDO_VAR3_PRIV:%.+]] = getelementptr [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], i64 [[OFFSET]] 1110 // CHECK: [[VAR3_PRIV:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[PSEUDO_VAR3_PRIV]] to [4 x [[S_FLOAT_TY]]]* 1111 1112 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 1113 1114 // CHECK: ret void 1115 1116 // CHECK: define internal void [[MAIN_MICROTASK10]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* dereferenceable(16) %{{.+}}) 1117 1118 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 1119 // CHECK: [[VAR3_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]], 1120 1121 // Reduction list for runtime. 1122 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1123 1124 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1125 1126 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1127 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1128 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1129 1130 // CHECK: [[LOW:%.+]] = getelementptr inbounds [4 x [[S_FLOAT_TY]]], [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], i64 0, i64 0 1131 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1132 1133 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1134 // CHECK: [[ORIG_START:%.+]] = bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 1135 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1136 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1137 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1138 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1139 // CHECK: [[PSEUDO_VAR3_PRIV:%.+]] = getelementptr [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], i64 [[OFFSET]] 1140 // CHECK: [[VAR3_PRIV:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[PSEUDO_VAR3_PRIV]] to [4 x [[S_FLOAT_TY]]]* 1141 1142 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 1143 1144 // CHECK: ret void 1145 1146 // CHECK: define internal void [[MAIN_MICROTASK11]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* dereferenceable(16) %{{.+}}) 1147 1148 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 1149 1150 // Reduction list for runtime. 1151 // CHECK: [[RED_LIST:%.+]] = alloca [2 x i8*], 1152 1153 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1154 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1155 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1156 1157 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64 1158 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW:%.+]] to i64 1159 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]] 1160 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1161 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1 1162 // CHECK: call i8* @llvm.stacksave() 1163 // CHECK: [[VAR3_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]], 1164 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1165 // CHECK: [[ORIG_START:%.+]] = bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 1166 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1167 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1168 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1169 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1170 // CHECK: [[PSEUDO_VAR3_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR3_PRIV]], i64 [[OFFSET]] 1171 // CHECK: [[VAR3_PRIV:%.+]] = bitcast [[S_FLOAT_TY]]* [[PSEUDO_VAR3_PRIV]] to [4 x [[S_FLOAT_TY]]]* 1172 1173 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 1174 1175 // CHECK: ret void 1176 1177 // CHECK: define internal void [[MAIN_MICROTASK12]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* dereferenceable(16) %{{.+}}) 1178 1179 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 1180 // CHECK: [[VAR3_PRIV:%.+]] = alloca [4 x [[S_FLOAT_TY]]], 1181 1182 // Reduction list for runtime. 1183 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1184 1185 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1186 1187 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1188 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1189 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1190 // CHECK: getelementptr inbounds [4 x [[S_FLOAT_TY]]], [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], i32 0, i32 0 1191 // CHECK: getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 4 1192 1193 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 1194 // CHECK: bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 1195 1196 // CHECK: ret void 1197 1198 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT_42]]() 1199 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]], 1200 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]]) 1201 // 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 1202 // Not interested in this one: 1203 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4, 1204 // 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 1205 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]* 1206 // CHECK: ret 1207 // 1208 // CHECK: define internal void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i32* dereferenceable(4) %{{.+}}, [[S_INT_TY]]* dereferenceable(4) %{{.+}}, [[S_INT_TY]]* dereferenceable(4) %{{.+}}, i32* dereferenceable(4) %{{.+}}, [2 x i32]* dereferenceable(8) %{{.+}}, [2 x [[S_INT_TY]]]* dereferenceable(8) %{{.+}}) 1209 // CHECK: alloca i{{[0-9]+}}, 1210 // CHECK: alloca i{{[0-9]+}}, 1211 // CHECK: alloca i{{[0-9]+}}, 1212 // CHECK: alloca i{{[0-9]+}}, 1213 // CHECK: alloca i{{[0-9]+}}, 1214 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 1215 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], 1216 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_INT_TY]], 1217 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca i{{[0-9]+}}, 1218 1219 // Reduction list for runtime. 1220 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 1221 1222 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1223 1224 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 1225 // CHECK: [[VAR1_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 1226 // CHECK: [[T_VAR1_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 1227 1228 // For + reduction operation initial value of private variable is 0. 1229 // CHECK: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[T_VAR_PRIV]], 1230 1231 // For & reduction operation initial value of private variable is ones in all bits. 1232 // CHECK: [[VAR_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 1233 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR_PRIV]]) 1234 1235 // For && reduction operation initial value of private variable is 1.0. 1236 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR1_PRIV]]) 1237 1238 // For min reduction operation initial value of private variable is largest repesentable value. 1239 // CHECK: store i{{[0-9]+}} 2147483647, i{{[0-9]+}}* [[T_VAR1_PRIV]], 1240 1241 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 1242 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 1243 // CHECK: call void @__kmpc_for_static_init_4( 1244 // Skip checks for internal operations. 1245 // CHECK: call void @__kmpc_for_static_fini( 1246 1247 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 1248 1249 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 1250 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR_PRIV]] to i8* 1251 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 1252 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 1253 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR_PRIV]] to i8* 1254 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 1255 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 1256 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_PRIV]] to i8* 1257 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 1258 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 1259 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR1_PRIV]] to i8* 1260 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 1261 1262 // res = __kmpc_reduce_nowait(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 1263 1264 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 1265 // 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]]) 1266 1267 // switch(res) 1268 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 1269 // CHECK: i32 1, label %[[CASE1:.+]] 1270 // CHECK: i32 2, label %[[CASE2:.+]] 1271 // CHECK: ] 1272 1273 // case 1: 1274 // t_var += t_var_reduction; 1275 // CHECK: [[T_VAR_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_REF]], 1276 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]], 1277 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_VAL]], [[T_VAR_PRIV_VAL]] 1278 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_REF]], 1279 1280 // var = var.operator &(var_reduction); 1281 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 1282 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* 1283 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 1284 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 1285 1286 // var1 = var1.operator &&(var1_reduction); 1287 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) 1288 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1289 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 1290 // CHECK: [[TRUE]] 1291 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) 1292 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1293 // CHECK: br label %[[END2]] 1294 // CHECK: [[END2]] 1295 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 1296 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 1297 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 1298 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* 1299 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 1300 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 1301 1302 // t_var1 = min(t_var1, t_var1_reduction); 1303 // CHECK: [[T_VAR1_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_REF]], 1304 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]], 1305 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_VAL]], [[T_VAR1_PRIV_VAL]] 1306 // CHECK: br i1 [[CMP]] 1307 // CHECK: [[UP:%.+]] = phi i32 1308 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_REF]], 1309 1310 // __kmpc_end_reduce_nowait(<loc>, <gtid>, &<lock>); 1311 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 1312 1313 // break; 1314 // CHECK: br label %[[RED_DONE]] 1315 1316 // case 2: 1317 // t_var += t_var_reduction; 1318 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]] 1319 // CHECK: atomicrmw add i32* [[T_VAR_REF]], i32 [[T_VAR_PRIV_VAL]] monotonic 1320 1321 // var = var.operator &(var_reduction); 1322 // CHECK: call void @__kmpc_critical( 1323 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 1324 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* 1325 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 1326 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 1327 // CHECK: call void @__kmpc_end_critical( 1328 1329 // var1 = var1.operator &&(var1_reduction); 1330 // CHECK: call void @__kmpc_critical( 1331 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) 1332 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1333 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 1334 // CHECK: [[TRUE]] 1335 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) 1336 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1337 // CHECK: br label %[[END2]] 1338 // CHECK: [[END2]] 1339 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 1340 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 1341 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 1342 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* 1343 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 1344 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 1345 // CHECK: call void @__kmpc_end_critical( 1346 1347 // t_var1 = min(t_var1, t_var1_reduction); 1348 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]] 1349 // CHECK: atomicrmw min i32* [[T_VAR1_REF]], i32 [[T_VAR1_PRIV_VAL]] monotonic 1350 1351 // break; 1352 // CHECK: br label %[[RED_DONE]] 1353 // CHECK: [[RED_DONE]] 1354 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) 1355 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* 1356 // CHECK: ret void 1357 1358 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 1359 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 1360 // ... 1361 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 1362 // *(Type<n>-1*)rhs[<n>-1]); 1363 // } 1364 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 1365 // t_var_lhs = (i{{[0-9]+}}*)lhs[0]; 1366 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 1367 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 1368 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to i{{[0-9]+}}* 1369 // t_var_rhs = (i{{[0-9]+}}*)rhs[0]; 1370 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 1371 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 1372 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to i{{[0-9]+}}* 1373 1374 // var_lhs = (S<i{{[0-9]+}}>*)lhs[1]; 1375 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 1376 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 1377 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_INT_TY]]* 1378 // var_rhs = (S<i{{[0-9]+}}>*)rhs[1]; 1379 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 1380 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 1381 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_INT_TY]]* 1382 1383 // var1_lhs = (S<i{{[0-9]+}}>*)lhs[2]; 1384 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 1385 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 1386 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_INT_TY]]* 1387 // var1_rhs = (S<i{{[0-9]+}}>*)rhs[2]; 1388 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 1389 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 1390 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_INT_TY]]* 1391 1392 // t_var1_lhs = (i{{[0-9]+}}*)lhs[3]; 1393 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 1394 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 1395 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to i{{[0-9]+}}* 1396 // t_var1_rhs = (i{{[0-9]+}}*)rhs[3]; 1397 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 1398 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 1399 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to i{{[0-9]+}}* 1400 1401 // t_var_lhs += t_var_rhs; 1402 // CHECK: [[T_VAR_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_LHS]], 1403 // CHECK: [[T_VAR_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_RHS]], 1404 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_LHS_VAL]], [[T_VAR_RHS_VAL]] 1405 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_LHS]], 1406 1407 // var_lhs = var_lhs.operator &(var_rhs); 1408 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_LHS]], [[S_INT_TY]]* dereferenceable(4) [[VAR_RHS]]) 1409 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_LHS]] to i8* 1410 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 1411 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 1412 1413 // var1_lhs = var1_lhs.operator &&(var1_rhs); 1414 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_LHS]]) 1415 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1416 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 1417 // CHECK: [[TRUE]] 1418 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_RHS]]) 1419 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1420 // CHECK: br label %[[END2]] 1421 // CHECK: [[END2]] 1422 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 1423 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 1424 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 1425 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_LHS]] to i8* 1426 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 1427 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) 1428 1429 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 1430 // CHECK: [[T_VAR1_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_LHS]], 1431 // CHECK: [[T_VAR1_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_RHS]], 1432 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_LHS_VAL]], [[T_VAR1_RHS_VAL]] 1433 // CHECK: br i1 [[CMP]] 1434 // CHECK: [[UP:%.+]] = phi i32 1435 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_LHS]], 1436 // CHECK: ret void 1437 1438 // CHECK: define internal void [[TMAIN_MICROTASK2]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [42 x [[S_INT_TY]]]* dereferenceable(168) %{{.*}}, [2 x i32]* dereferenceable(8) %{{.*}}, i32* dereferenceable(4) %{{.*}}, [2 x [[S_INT_TY]]]* dereferenceable(8) %{{.*}}, [[S_INT_TY]]* dereferenceable(4) %{{.*}}) 1439 1440 // CHECK: [[ARR_ORIG_ADDR:%.+]] = alloca [42 x [[S_INT_TY]]]*, 1441 // CHECK: [[ARR_PRIV:%.+]] = alloca [40 x [[S_INT_TY]]], 1442 1443 // Reduction list for runtime. 1444 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1445 1446 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1447 1448 // CHECK: [[ARR_ORIG:%.+]] = load [42 x [[S_INT_TY]]]*, [42 x [[S_INT_TY]]]** [[ARR_ORIG_ADDR]], 1449 // CHECK: [[LOW:%.+]] = getelementptr inbounds [42 x [[S_INT_TY]]], [42 x [[S_INT_TY]]]* [[ARR_ORIG]], i64 0, i64 1 1450 // CHECK: [[ORIG_START:%.+]] = bitcast [42 x [[S_INT_TY]]]* [[ARR_ORIG]] to [[S_INT_TY]]* 1451 // CHECK: [[START:%.+]] = ptrtoint [[S_INT_TY]]* [[ORIG_START]] to i64 1452 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_INT_TY]]* [[LOW]] to i64 1453 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1454 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64) 1455 // CHECK: [[PSEUDO_ARR_PRIV:%.+]] = getelementptr [40 x [[S_INT_TY]]], [40 x [[S_INT_TY]]]* [[ARR_PRIV]], i64 [[OFFSET]] 1456 // CHECK: [[ARR_PRIV:%.+]] = bitcast [40 x [[S_INT_TY]]]* [[PSEUDO_ARR_PRIV]] to [42 x [[S_INT_TY]]]* 1457 1458 // CHECK: ret void 1459 1460 #endif 1461 1462