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 global %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* 33 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr global %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8* 34 // CHECK-DAG: [[REDUCTION_LOC:@.+]] = private unnamed_addr global %{{.+}} { 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* 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) %{{.+}}) 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: [[VAR1_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** % 296 // CHECK: [[T_VAR1_REF:%.+]] = load float*, float** % 297 298 // For + reduction operation initial value of private variable is 0. 299 // CHECK: store float 0.0{{.+}}, float* [[T_VAR_PRIV]], 300 301 // For & reduction operation initial value of private variable is ones in all bits. 302 // CHECK: [[VAR_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** % 303 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 304 305 // For && reduction operation initial value of private variable is 1.0. 306 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[VAR1_PRIV]]) 307 308 // For min reduction operation initial value of private variable is largest repesentable value. 309 // CHECK: store float 0x47EFFFFFE0000000, float* [[T_VAR1_PRIV]], 310 311 312 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 313 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 314 // CHECK: call void @__kmpc_for_static_init_4( 315 // Skip checks for internal operations. 316 // CHECK: call void @__kmpc_for_static_fini( 317 318 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 319 320 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 321 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR_PRIV]] to i8* 322 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 323 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 324 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_PRIV]] to i8* 325 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 326 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 327 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_PRIV]] to i8* 328 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 329 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 330 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR1_PRIV]] to i8* 331 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 332 333 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 334 335 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 336 // 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]]) 337 338 // switch(res) 339 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 340 // CHECK: i32 1, label %[[CASE1:.+]] 341 // CHECK: i32 2, label %[[CASE2:.+]] 342 // CHECK: ] 343 344 // case 1: 345 // t_var += t_var_reduction; 346 // CHECK: [[T_VAR_VAL:%.+]] = load float, float* [[T_VAR_REF]], 347 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load float, float* [[T_VAR_PRIV]], 348 // CHECK: [[UP:%.+]] = fadd float [[T_VAR_VAL]], [[T_VAR_PRIV_VAL]] 349 // CHECK: store float [[UP]], float* [[T_VAR_REF]], 350 351 // var = var.operator &(var_reduction); 352 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @{{.+}}([[S_FLOAT_TY]]* [[VAR_REF]], [[S_FLOAT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 353 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_REF]] to i8* 354 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[UP]] to i8* 355 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 356 357 // var1 = var1.operator &&(var1_reduction); 358 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_REF]]) 359 // CHECK: [[VAR1_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 360 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 361 // CHECK: [[TRUE]] 362 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_PRIV]]) 363 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 364 // CHECK: br label %[[END2]] 365 // CHECK: [[END2]] 366 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 367 // CHECK: [[CONV:%.+]] = uitofp i1 [[COND_LVALUE]] to float 368 // CHECK: call void @{{.+}}([[S_FLOAT_TY]]* [[COND_LVALUE:%.+]], float [[CONV]]) 369 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_REF]] to i8* 370 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[COND_LVALUE]] to i8* 371 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 372 373 // t_var1 = min(t_var1, t_var1_reduction); 374 // CHECK: [[T_VAR1_VAL:%.+]] = load float, float* [[T_VAR1_REF]], 375 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load float, float* [[T_VAR1_PRIV]], 376 // CHECK: [[CMP:%.+]] = fcmp olt float [[T_VAR1_VAL]], [[T_VAR1_PRIV_VAL]] 377 // CHECK: br i1 [[CMP]] 378 // CHECK: [[UP:%.+]] = phi float 379 // CHECK: store float [[UP]], float* [[T_VAR1_REF]], 380 381 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 382 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 383 384 // break; 385 // CHECK: br label %[[RED_DONE]] 386 387 // case 2: 388 // t_var += t_var_reduction; 389 // CHECK: load float, float* [[T_VAR_PRIV]] 390 // CHECK: [[T_VAR_REF_INT:%.+]] = bitcast float* [[T_VAR_REF]] to i32* 391 // CHECK: [[OLD1:%.+]] = load atomic i32, i32* [[T_VAR_REF_INT]] monotonic, 392 // CHECK: br label %[[CONT:.+]] 393 // CHECK: [[CONT]] 394 // CHECK: [[ORIG_OLD_INT:%.+]] = phi i32 [ [[OLD1]], %{{.+}} ], [ [[OLD2:%.+]], %[[CONT]] ] 395 // CHECK: fadd float 396 // CHECK: [[UP_INT:%.+]] = load i32, i32* 397 // CHECK: [[T_VAR_REF_INT:%.+]] = bitcast float* [[T_VAR_REF]] to i32* 398 // CHECK: [[RES:%.+]] = cmpxchg i32* [[T_VAR_REF_INT]], i32 [[ORIG_OLD_INT]], i32 [[UP_INT]] monotonic monotonic 399 // CHECK: [[OLD2:%.+]] = extractvalue { i32, i1 } [[RES]], 0 400 // CHECK: [[SUCCESS_FAIL:%.+]] = extractvalue { i32, i1 } [[RES]], 1 401 // CHECK: br i1 [[SUCCESS_FAIL]], label %[[ATOMIC_DONE:.+]], label %[[CONT]] 402 // CHECK: [[ATOMIC_DONE]] 403 404 // var = var.operator &(var_reduction); 405 // CHECK: call void @__kmpc_critical( 406 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @{{.+}}([[S_FLOAT_TY]]* [[VAR_REF]], [[S_FLOAT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 407 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_REF]] to i8* 408 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[UP]] to i8* 409 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 410 // CHECK: call void @__kmpc_end_critical( 411 412 // var1 = var1.operator &&(var1_reduction); 413 // CHECK: call void @__kmpc_critical( 414 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_REF]]) 415 // CHECK: [[VAR1_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 416 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 417 // CHECK: [[TRUE]] 418 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_PRIV]]) 419 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 420 // CHECK: br label %[[END2]] 421 // CHECK: [[END2]] 422 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 423 // CHECK: [[CONV:%.+]] = uitofp i1 [[COND_LVALUE]] to float 424 // CHECK: call void @{{.+}}([[S_FLOAT_TY]]* [[COND_LVALUE:%.+]], float [[CONV]]) 425 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_REF]] to i8* 426 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[COND_LVALUE]] to i8* 427 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 428 // CHECK: call void @__kmpc_end_critical( 429 430 // t_var1 = min(t_var1, t_var1_reduction); 431 // CHECK: load float, float* [[T_VAR1_PRIV]] 432 // CHECK: [[T_VAR1_REF_INT:%.+]] = bitcast float* [[T_VAR1_REF]] to i32* 433 // CHECK: [[OLD1:%.+]] = load atomic i32, i32* [[T_VAR1_REF_INT]] monotonic, 434 // CHECK: br label %[[CONT:.+]] 435 // CHECK: [[CONT]] 436 // CHECK: [[ORIG_OLD_INT:%.+]] = phi i32 [ [[OLD1]], %{{.+}} ], [ [[OLD2:%.+]], %{{.+}} ] 437 // CHECK: [[CMP:%.+]] = fcmp olt float 438 // CHECK: br i1 [[CMP]] 439 // CHECK: phi float 440 // CHECK: [[UP_INT:%.+]] = load i32 441 // CHECK: [[T_VAR1_REF_INT:%.+]] = bitcast float* [[T_VAR1_REF]] to i32* 442 // CHECK: [[RES:%.+]] = cmpxchg i32* [[T_VAR1_REF_INT]], i32 [[ORIG_OLD_INT]], i32 [[UP_INT]] monotonic monotonic 443 // CHECK: [[OLD2:%.+]] = extractvalue { i32, i1 } [[RES]], 0 444 // CHECK: [[SUCCESS_FAIL:%.+]] = extractvalue { i32, i1 } [[RES]], 1 445 // CHECK: br i1 [[SUCCESS_FAIL]], label %[[ATOMIC_DONE:.+]], label %[[CONT]] 446 // CHECK: [[ATOMIC_DONE]] 447 448 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 449 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 450 451 // break; 452 // CHECK: br label %[[RED_DONE]] 453 // CHECK: [[RED_DONE]] 454 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 455 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* 456 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) 457 458 // CHECK: ret void 459 460 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 461 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 462 // ... 463 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 464 // *(Type<n>-1*)rhs[<n>-1]); 465 // } 466 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 467 // t_var_lhs = (float*)lhs[0]; 468 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 469 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 470 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to float* 471 // t_var_rhs = (float*)rhs[0]; 472 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 473 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 474 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to float* 475 476 // var_lhs = (S<float>*)lhs[1]; 477 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 478 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 479 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_FLOAT_TY]]* 480 // var_rhs = (S<float>*)rhs[1]; 481 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 482 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 483 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_FLOAT_TY]]* 484 485 // var1_lhs = (S<float>*)lhs[2]; 486 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 487 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 488 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_FLOAT_TY]]* 489 // var1_rhs = (S<float>*)rhs[2]; 490 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 491 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 492 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_FLOAT_TY]]* 493 494 // t_var1_lhs = (float*)lhs[3]; 495 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 496 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 497 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to float* 498 // t_var1_rhs = (float*)rhs[3]; 499 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 500 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 501 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to float* 502 503 // t_var_lhs += t_var_rhs; 504 // CHECK: [[T_VAR_LHS_VAL:%.+]] = load float, float* [[T_VAR_LHS]], 505 // CHECK: [[T_VAR_RHS_VAL:%.+]] = load float, float* [[T_VAR_RHS]], 506 // CHECK: [[UP:%.+]] = fadd float [[T_VAR_LHS_VAL]], [[T_VAR_RHS_VAL]] 507 // CHECK: store float [[UP]], float* [[T_VAR_LHS]], 508 509 // var_lhs = var_lhs.operator &(var_rhs); 510 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @{{.+}}([[S_FLOAT_TY]]* [[VAR_LHS]], [[S_FLOAT_TY]]* dereferenceable(4) [[VAR_RHS]]) 511 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_LHS]] to i8* 512 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[UP]] to i8* 513 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 514 515 // var1_lhs = var1_lhs.operator &&(var1_rhs); 516 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_LHS]]) 517 // CHECK: [[VAR1_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 518 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 519 // CHECK: [[TRUE]] 520 // CHECK: [[TO_FLOAT:%.+]] = call float @{{.+}}([[S_FLOAT_TY]]* [[VAR1_RHS]]) 521 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = fcmp une float [[TO_FLOAT]], 0.0 522 // CHECK: br label %[[END2]] 523 // CHECK: [[END2]] 524 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 525 // CHECK: [[CONV:%.+]] = uitofp i1 [[COND_LVALUE]] to float 526 // CHECK: call void @{{.+}}([[S_FLOAT_TY]]* [[COND_LVALUE:%.+]], float [[CONV]]) 527 // CHECK: [[BC1:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_LHS]] to i8* 528 // CHECK: [[BC2:%.+]] = bitcast [[S_FLOAT_TY]]* [[COND_LVALUE]] to i8* 529 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 530 531 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 532 // CHECK: [[T_VAR1_LHS_VAL:%.+]] = load float, float* [[T_VAR1_LHS]], 533 // CHECK: [[T_VAR1_RHS_VAL:%.+]] = load float, float* [[T_VAR1_RHS]], 534 // CHECK: [[CMP:%.+]] = fcmp olt float [[T_VAR1_LHS_VAL]], [[T_VAR1_RHS_VAL]] 535 // CHECK: br i1 [[CMP]] 536 // CHECK: [[UP:%.+]] = phi float 537 // CHECK: store float [[UP]], float* [[T_VAR1_LHS]], 538 // CHECK: ret void 539 540 // 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) %{{.+}}) 541 542 // Reduction list for runtime. 543 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 544 545 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 546 // CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}** 547 // CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}** [[ARR_ADDR:%.+]], 548 // CHECK: [[ARR:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[ARR_ADDR]], 549 550 // CHECK: [[UB_CAST:%.+]] = ptrtoint i32* [[UB1_UP:%.+]] to i64 551 // CHECK: [[LB_CAST:%.+]] = ptrtoint i32* [[LB1_0:%.+]] to i64 552 // CHECK: [[DIFF:%.+]] = sub i64 [[UB_CAST]], [[LB_CAST]] 553 // CHECK: [[SIZE_1:%.+]] = sdiv exact i64 [[DIFF]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64) 554 // CHECK: [[ARR_SIZE:%.+]] = add nuw i64 [[SIZE_1]], 1 555 // CHECK: call i8* @llvm.stacksave() 556 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]], 557 558 // Check initialization of private copy. 559 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]] 560 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]] 561 // CHECK: br i1 [[ISEMPTY]], 562 // CHECK: phi i32* 563 // CHECK: store i32 0, i32* % 564 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 565 // CHECK: br i1 [[DONE]], 566 567 // CHECK: [[ARRS_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[ARRS_SIZE:%.+]], 568 569 // Check initialization of private copy. 570 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]] 571 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]] 572 // CHECK: br i1 [[ISEMPTY]], 573 // CHECK: phi [[S_FLOAT_TY]]* 574 // CHECK: call void @_ZN1SIfEC1Ev([[S_FLOAT_TY]]* % 575 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 576 // CHECK: br i1 [[DONE]], 577 578 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 579 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 580 // CHECK: call void @__kmpc_for_static_init_4( 581 // Skip checks for internal operations. 582 // CHECK: call void @__kmpc_for_static_fini( 583 584 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 585 586 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 587 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8* 588 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]], 589 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 590 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8* 591 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]], 592 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 593 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV]] to i8* 594 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]], 595 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 596 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARRS_SIZE]] to i8* 597 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_SIZE_REF]], 598 599 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 600 601 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 602 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 603 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 604 // 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]]) 605 606 // switch(res) 607 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 608 // CHECK: i32 1, label %[[CASE1:.+]] 609 // CHECK: i32 2, label %[[CASE2:.+]] 610 // CHECK: ] 611 612 // case 1: 613 // CHECK: [[CASE1]] 614 615 // arr[:] += arr_reduction[:]; 616 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 617 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 618 // CHECK: br i1 [[ISEMPTY]], 619 // CHECK: phi i32* 620 // CHECK: [[ADD:%.+]] = add nsw i32 % 621 // CHECK: store i32 [[ADD]], i32* % 622 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 623 // CHECK: br i1 [[DONE]], 624 625 // arrs[:] = var.operator &(arrs_reduction[:]); 626 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 627 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 628 // CHECK: br i1 [[ISEMPTY]], 629 // CHECK: phi [[S_FLOAT_TY]]* 630 // CHECK: [[AND:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}) 631 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 632 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %{{.+}}, i8* align 4 [[BITCAST]], i64 4, i1 false) 633 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 634 // CHECK: br i1 [[DONE]], 635 636 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 637 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 638 639 // break; 640 // CHECK: br label %[[RED_DONE]] 641 642 // case 2: 643 // CHECK: [[CASE2]] 644 645 // arr[:] += arr_reduction[:]; 646 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 647 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 648 // CHECK: br i1 [[ISEMPTY]], 649 // CHECK: phi i32* 650 // CHECK: atomicrmw add i32* %{{.+}}, i32 %{{.+}} monotonic 651 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 652 // CHECK: br i1 [[DONE]], 653 654 // arrs[:] = var.operator &(arrs_reduction[:]); 655 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 656 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 657 // CHECK: br i1 [[ISEMPTY]], 658 // CHECK: phi [[S_FLOAT_TY]]* 659 // CHECK: call void @__kmpc_critical( 660 // CHECK: [[AND:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}) 661 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 662 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %{{.+}}, i8* align 4 [[BITCAST]], i64 4, i1 false) 663 // CHECK: call void @__kmpc_end_critical( 664 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 665 // CHECK: br i1 [[DONE]], 666 667 // break; 668 // CHECK: br label %[[RED_DONE]] 669 // CHECK: [[RED_DONE]] 670 671 // Check destruction of private copy. 672 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]] 673 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]] 674 // CHECK: br i1 [[ISEMPTY]], 675 // CHECK: phi [[S_FLOAT_TY]]* 676 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* % 677 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[ARRS_PRIV]] 678 // CHECK: br i1 [[DONE]], 679 // CHECK: call void @llvm.stackrestore(i8* 680 681 // CHECK: ret void 682 683 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 684 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 685 // ... 686 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 687 // *(Type<n>-1*)rhs[<n>-1]); 688 // } 689 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 690 // arr_rhs = (int*)rhs[0]; 691 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 692 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]], 693 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32* 694 // arr_lhs = (int*)lhs[0]; 695 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 696 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]], 697 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32* 698 699 // arr_size = (size_t)lhs[1]; 700 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 701 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]], 702 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64 703 704 // arrs_rhs = (S<float>*)rhs[2]; 705 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 706 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]], 707 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]* 708 // arrs_lhs = (S<float>*)lhs[2]; 709 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 710 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]], 711 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]* 712 713 // arrs_size = (size_t)lhs[3]; 714 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 715 // CHECK: [[ARRS_SIZE_VOID:%.+]] = load i8*, i8** [[ARRS_SIZE_REF]], 716 // CHECK: [[ARRS_SIZE:%.+]] = ptrtoint i8* [[ARRS_SIZE_VOID]] to i64 717 718 // arr_lhs[:] += arr_rhs[:]; 719 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]] 720 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]] 721 // CHECK: br i1 [[ISEMPTY]], 722 // CHECK: phi i32* 723 // CHECK: [[ADD:%.+]] = add nsw i32 % 724 // CHECK: store i32 [[ADD]], i32* % 725 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 726 // CHECK: br i1 [[DONE]], 727 728 // arrs_lhs = arrs_lhs.operator &(arrs_rhs); 729 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]] 730 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 731 // CHECK: br i1 [[ISEMPTY]], 732 // CHECK: phi [[S_FLOAT_TY]]* 733 // CHECK: [[AND:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}) 734 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 735 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %{{.+}}, i8* align 4 [[BITCAST]], i64 4, i1 false) 736 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 737 // CHECK: br i1 [[DONE]], 738 739 // CHECK: ret void 740 741 // 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) %{{.+}}) 742 743 // CHECK: [[ARRS_PRIV:%.+]] = alloca [10 x [4 x [[S_FLOAT_TY]]]], 744 745 // Reduction list for runtime. 746 // CHECK: [[RED_LIST:%.+]] = alloca [3 x i8*], 747 748 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 749 750 // CHECK: [[ARR_SIZE:%.+]] = udiv exact i64 751 // CHECK: call i8* @llvm.stacksave() 752 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]], 753 754 // Check initialization of private copy. 755 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]] 756 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]] 757 // CHECK: br i1 [[ISEMPTY]], 758 // CHECK: phi i32* 759 // CHECK: store i32 0, i32* % 760 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 761 // CHECK: br i1 [[DONE]], 762 763 // Check initialization of private copy. 764 // 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 0 765 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40 766 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[BEGIN]], [[END]] 767 // CHECK: br i1 [[ISEMPTY]], 768 // CHECK: phi [[S_FLOAT_TY]]* 769 // CHECK: call void @_ZN1SIfEC1Ev([[S_FLOAT_TY]]* % 770 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 771 // CHECK: br i1 [[DONE]], 772 // CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]* 773 // CHECK: [[ARRS_PRIV_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]] to [[S_FLOAT_TY]]* 774 775 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 776 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 777 // CHECK: call void @__kmpc_for_static_init_4( 778 // Skip checks for internal operations. 779 // CHECK: call void @__kmpc_for_static_fini( 780 781 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 782 783 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 0 784 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8* 785 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]], 786 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 1 787 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8* 788 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]], 789 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 2 790 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV_BEGIN]] to i8* 791 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]], 792 793 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 794 795 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 796 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 797 // CHECK: [[BITCAST:%.+]] = bitcast [3 x i8*]* [[RED_LIST]] to i8* 798 // 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]]) 799 800 // switch(res) 801 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 802 // CHECK: i32 1, label %[[CASE1:.+]] 803 // CHECK: i32 2, label %[[CASE2:.+]] 804 // CHECK: ] 805 806 // case 1: 807 // CHECK: [[CASE1]] 808 809 // arr[:] += arr_reduction[:]; 810 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0:%.+]], i64 [[ARR_SIZE]] 811 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 812 // CHECK: br i1 [[ISEMPTY]], 813 // CHECK: phi i32* 814 // CHECK: [[ADD:%.+]] = add nsw i32 % 815 // CHECK: store i32 [[ADD]], i32* % 816 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 817 // CHECK: br i1 [[DONE]], 818 819 // arrs[:] = var.operator &(arrs_reduction[:]); 820 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40 821 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]] 822 // CHECK: br i1 [[ISEMPTY]], 823 // CHECK: phi [[S_FLOAT_TY]]* 824 // CHECK: [[AND:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}) 825 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 826 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %{{.+}}, i8* align 4 [[BITCAST]], i64 4, i1 false) 827 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 828 // CHECK: br i1 [[DONE]], 829 830 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>); 831 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 832 833 // break; 834 // CHECK: br label %[[RED_DONE]] 835 836 // case 2: 837 // CHECK: [[CASE2]] 838 839 // arr[:] += arr_reduction[:]; 840 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]] 841 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]] 842 // CHECK: br i1 [[ISEMPTY]], 843 // CHECK: phi i32* 844 // CHECK: atomicrmw add i32* %{{.+}}, i32 %{{.+}} monotonic 845 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 846 // CHECK: br i1 [[DONE]], 847 848 // arrs[:] = var.operator &(arrs_reduction[:]); 849 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40 850 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]] 851 // CHECK: br i1 [[ISEMPTY]], 852 // CHECK: phi [[S_FLOAT_TY]]* 853 // CHECK: call void @__kmpc_critical( 854 // CHECK: [[AND:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}) 855 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 856 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %{{.+}}, i8* align 4 [[BITCAST]], i64 4, i1 false) 857 // CHECK: call void @__kmpc_end_critical( 858 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 859 // CHECK: br i1 [[DONE]], 860 861 // break; 862 // CHECK: br label %[[RED_DONE]] 863 // CHECK: [[RED_DONE]] 864 865 // Check destruction of private copy. 866 // 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 0 867 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40 868 // CHECK: br 869 // CHECK: phi [[S_FLOAT_TY]]* 870 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* % 871 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[BEGIN]] 872 // CHECK: br i1 [[DONE]], 873 // CHECK: call void @llvm.stackrestore(i8* 874 // CHECK: call void @__kmpc_barrier( 875 876 // CHECK: ret void 877 878 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 879 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 880 // ... 881 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 882 // *(Type<n>-1*)rhs[<n>-1]); 883 // } 884 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 885 // arr_rhs = (int*)rhs[0]; 886 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 887 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]], 888 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32* 889 // arr_lhs = (int*)lhs[0]; 890 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 891 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]], 892 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32* 893 894 // arr_size = (size_t)lhs[1]; 895 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 896 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]], 897 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64 898 899 // arrs_rhs = (S<float>*)rhs[2]; 900 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 901 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]], 902 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]* 903 // arrs_lhs = (S<float>*)lhs[2]; 904 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 905 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]], 906 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]* 907 908 // arr_lhs[:] += arr_rhs[:]; 909 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]] 910 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]] 911 // CHECK: br i1 [[ISEMPTY]], 912 // CHECK: phi i32* 913 // CHECK: [[ADD:%.+]] = add nsw i32 % 914 // CHECK: store i32 [[ADD]], i32* % 915 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]] 916 // CHECK: br i1 [[DONE]], 917 918 // arrs_lhs = arrs_lhs.operator &(arrs_rhs); 919 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 40 920 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]] 921 // CHECK: br i1 [[ISEMPTY]], 922 // CHECK: phi [[S_FLOAT_TY]]* 923 // CHECK: [[AND:%.+]] = call dereferenceable(4) [[S_FLOAT_TY]]* @_ZN1SIfEanERKS0_([[S_FLOAT_TY]]* %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}) 924 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[AND]] to i8* 925 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %{{.+}}, i8* align 4 [[BITCAST]], i64 4, i1 false) 926 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]] 927 // CHECK: br i1 [[DONE]], 928 929 // CHECK: ret void 930 931 // CHECK: define internal void [[MAIN_MICROTASK3]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i32* {{.*}} %{{.+}}) 932 933 // CHECK: [[VLA1_ORIG_ADDR:%.+]] = alloca i64 934 // CHECK: [[VLA2_ORIG_ADDR:%.+]] = alloca i64 935 // CHECK: [[ARR_ORIG_ADDR:%.+]] = alloca i32*, 936 // CHECK: [[ARR_PRIV:%.+]] = alloca [1 x [2 x i32]], 937 938 // Reduction list for runtime. 939 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 940 941 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 942 // CHECK: [[VLA1:%.+]] = load i64, i64* [[VLA1_ORIG_ADDR]] 943 // CHECK: [[VLA2:%.+]] = load i64, i64* [[VLA2_ORIG_ADDR]] 944 // CHECK: [[ARR_ORIG:%.+]] = load i32*, i32** [[ARR_ORIG_ADDR]], 945 946 // CHECK: [[LOW_OFFSET:%.+]] = mul nsw i64 1, [[VLA2]] 947 // CHECK: [[ARRIDX:%.+]] = getelementptr inbounds i32, i32* [[ARR_ORIG]], i64 [[LOW_OFFSET]] 948 // CHECK: [[LOW:%.+]] = getelementptr inbounds i32, i32* [[ARRIDX]], i64 0 949 950 // CHECK: [[START:%.+]] = ptrtoint i32* [[ARR_ORIG]] to i64 951 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint i32* [[LOW]] to i64 952 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 953 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64) 954 // CHECK: [[ARR_PRIV_PTR:%.+]] = bitcast [1 x [2 x i32]]* [[ARR_PRIV]] to i32* 955 // CHECK: [[ARR_PRIV:%.+]] = getelementptr i32, i32* [[ARR_PRIV_PTR]], i64 [[OFFSET]] 956 957 // CHECK: ret void 958 959 // CHECK: define internal void [[MAIN_MICROTASK4]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** dereferenceable(8) %{{.+}}) 960 961 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***, 962 963 // Reduction list for runtime. 964 // CHECK: [[RED_LIST:%.+]] = alloca [2 x i8*], 965 966 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 967 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]], 968 969 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64 970 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW:%.+]] to i64 971 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]] 972 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 973 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1 974 // CHECK: call i8* @llvm.stacksave() 975 // CHECK: [[VAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]], 976 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 977 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], 978 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 979 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 980 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 981 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 982 // CHECK: [[PSEUDO_VAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR2_PRIV]], i64 [[OFFSET]] 983 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** % 984 // CHECK: store [[S_FLOAT_TY]]* [[PSEUDO_VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]] 985 // CHECK: ret void 986 987 // CHECK: define internal void [[MAIN_MICROTASK5]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** dereferenceable(8) %{{.+}}) 988 989 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***, 990 // CHECK: [[VAR2_PRIV:%.+]] = alloca [1 x [6 x [[S_FLOAT_TY]]]], 991 992 // Reduction list for runtime. 993 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 994 995 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 996 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]], 997 998 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 999 // CHECK: [[ARRIDX:%.+]] = getelementptr inbounds [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], i64 1 1000 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[ARRIDX]], 1001 // CHECK: [[LOW:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LD]], i64 1 1002 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1003 1004 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1005 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], 1006 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1007 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1008 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1009 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1010 // CHECK: [[VAR2_PRIV_PTR:%.+]] = bitcast [1 x [6 x [[S_FLOAT_TY]]]]* [[VAR2_PRIV]] to [[S_FLOAT_TY]]* 1011 // CHECK: [[VAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR2_PRIV_PTR]], i64 [[OFFSET]] 1012 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** % 1013 // CHECK: store [[S_FLOAT_TY]]* [[VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]] 1014 // CHECK: ret void 1015 1016 // CHECK: define internal void [[MAIN_MICROTASK6]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** dereferenceable(8) %{{.+}}) 1017 1018 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***, 1019 // CHECK: [[VAR2_PRIV:%.+]] = alloca [1 x [6 x [[S_FLOAT_TY]]]], 1020 1021 // Reduction list for runtime. 1022 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1023 1024 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1025 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]], 1026 1027 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1028 // CHECK: [[ARRIDX:%.+]] = getelementptr inbounds [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], i64 1 1029 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[ARRIDX]], 1030 // CHECK: [[LOW:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LD]], i64 1 1031 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1032 1033 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1034 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], 1035 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1036 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1037 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1038 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1039 // CHECK: [[VAR2_PRIV_PTR:%.+]] = bitcast [1 x [6 x [[S_FLOAT_TY]]]]* [[VAR2_PRIV]] to [[S_FLOAT_TY]]* 1040 // CHECK: [[VAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR2_PRIV_PTR]], i64 [[OFFSET]] 1041 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** % 1042 // CHECK: store [[S_FLOAT_TY]]* [[VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]] 1043 // CHECK: ret void 1044 1045 // CHECK: define internal void [[MAIN_MICROTASK7]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** dereferenceable(8) %{{.+}}) 1046 1047 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***, 1048 // CHECK: [[VAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 1049 1050 // Reduction list for runtime. 1051 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1052 1053 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1054 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]], 1055 1056 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1057 // CHECK: [[ARRIDX:%.+]] = getelementptr inbounds [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], i64 1 1058 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[ARRIDX]], 1059 // CHECK: [[LOW:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LD]], i64 1 1060 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1061 1062 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]], 1063 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]], 1064 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1065 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1066 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1067 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1068 // CHECK: [[PSEUDO_VAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR2_PRIV]], i64 [[OFFSET]] 1069 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** % 1070 // CHECK: store [[S_FLOAT_TY]]* [[PSEUDO_VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]] 1071 // CHECK: ret void 1072 1073 // CHECK: define internal void [[MAIN_MICROTASK8]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [5 x [[S_FLOAT_TY]]]* dereferenceable(20) %{{.+}}) 1074 1075 // CHECK: [[VVAR2_ORIG_ADDR:%.+]] = alloca [5 x [[S_FLOAT_TY]]]*, 1076 // CHECK: [[VVAR2_PRIV:%.+]] = alloca [5 x [[S_FLOAT_TY]]], 1077 1078 // Reduction list for runtime. 1079 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1080 1081 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1082 // CHECK: [[VVAR2_ORIG:%.+]] = load [5 x [[S_FLOAT_TY]]]*, [5 x [[S_FLOAT_TY]]]** [[VVAR2_ORIG_ADDR]], 1083 1084 // CHECK: [[LOW:%.+]] = getelementptr inbounds [5 x [[S_FLOAT_TY]]], [5 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]], i64 0, i64 0 1085 // CHECK: [[ORIG_START:%.+]] = bitcast [5 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]] to [[S_FLOAT_TY]]* 1086 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1087 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1088 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1089 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1090 // CHECK: [[VVAR2_PRIV_PTR:%.+]] = bitcast [5 x [[S_FLOAT_TY]]]* [[VVAR2_PRIV]] to [[S_FLOAT_TY]]* 1091 // CHECK: [[VVAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VVAR2_PRIV_PTR]], i64 [[OFFSET]] 1092 // CHECK: ret void 1093 1094 // CHECK: define internal void [[MAIN_MICROTASK9]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* dereferenceable(16) %{{.+}}) 1095 1096 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 1097 // CHECK: [[VAR3_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]], 1098 1099 // Reduction list for runtime. 1100 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1101 1102 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1103 1104 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1105 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1106 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1107 1108 // CHECK: [[LOW:%.+]] = getelementptr inbounds [4 x [[S_FLOAT_TY]]], [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], i64 0, i64 1 1109 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1110 1111 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1112 // CHECK: [[ORIG_START:%.+]] = bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 1113 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1114 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1115 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1116 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1117 // CHECK: [[VAR3_PRIV_PTR:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]] to [[S_FLOAT_TY]]* 1118 // CHECK: [[PSEUDO_VAR3_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR3_PRIV_PTR]], i64 [[OFFSET]] 1119 // CHECK: [[VAR3_PRIV:%.+]] = bitcast [[S_FLOAT_TY]]* [[PSEUDO_VAR3_PRIV]] to [4 x [[S_FLOAT_TY]]]* 1120 1121 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 1122 1123 // CHECK: ret void 1124 1125 // CHECK: define internal void [[MAIN_MICROTASK10]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* dereferenceable(16) %{{.+}}) 1126 1127 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 1128 // CHECK: [[VAR3_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]], 1129 1130 // Reduction list for runtime. 1131 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1132 1133 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1134 1135 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1136 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1137 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1138 1139 // CHECK: [[LOW:%.+]] = getelementptr inbounds [4 x [[S_FLOAT_TY]]], [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], i64 0, i64 0 1140 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1141 1142 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1143 // CHECK: [[ORIG_START:%.+]] = bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 1144 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1145 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1146 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1147 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1148 // CHECK: [[VAR3_PRIV_PTR:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]] to [[S_FLOAT_TY]]* 1149 // CHECK: [[PSEUDO_VAR3_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR3_PRIV_PTR]], i64 [[OFFSET]] 1150 // CHECK: [[VAR3_PRIV:%.+]] = bitcast [[S_FLOAT_TY]]* [[PSEUDO_VAR3_PRIV]] to [4 x [[S_FLOAT_TY]]]* 1151 1152 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 1153 1154 // CHECK: ret void 1155 1156 // CHECK: define internal void [[MAIN_MICROTASK11]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* dereferenceable(16) %{{.+}}) 1157 1158 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 1159 1160 // Reduction list for runtime. 1161 // CHECK: [[RED_LIST:%.+]] = alloca [2 x i8*], 1162 1163 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1164 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1165 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1166 1167 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64 1168 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW:%.+]] to i64 1169 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]] 1170 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1171 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1 1172 // CHECK: call i8* @llvm.stacksave() 1173 // CHECK: [[VAR3_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]], 1174 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1175 // CHECK: [[ORIG_START:%.+]] = bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 1176 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64 1177 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64 1178 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1179 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 1180 // CHECK: [[PSEUDO_VAR3_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR3_PRIV]], i64 [[OFFSET]] 1181 // CHECK: [[VAR3_PRIV:%.+]] = bitcast [[S_FLOAT_TY]]* [[PSEUDO_VAR3_PRIV]] to [4 x [[S_FLOAT_TY]]]* 1182 1183 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 1184 1185 // CHECK: ret void 1186 1187 // CHECK: define internal void [[MAIN_MICROTASK12]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* dereferenceable(16) %{{.+}}) 1188 1189 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*, 1190 // CHECK: [[VAR3_PRIV:%.+]] = alloca [4 x [[S_FLOAT_TY]]], 1191 1192 // Reduction list for runtime. 1193 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1194 1195 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1196 1197 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1198 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]], 1199 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]], 1200 // CHECK: getelementptr inbounds [4 x [[S_FLOAT_TY]]], [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], i32 0, i32 0 1201 // CHECK: getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 4 1202 1203 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** % 1204 // CHECK: bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]* 1205 1206 // CHECK: ret void 1207 1208 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT_42]]() 1209 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]], 1210 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]]) 1211 // 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 1212 // Not interested in this one: 1213 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4, 1214 // 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 1215 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]* 1216 // CHECK: ret 1217 // 1218 // 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) %{{.+}}) 1219 // CHECK: alloca i{{[0-9]+}}, 1220 // CHECK: alloca i{{[0-9]+}}, 1221 // CHECK: alloca i{{[0-9]+}}, 1222 // CHECK: alloca i{{[0-9]+}}, 1223 // CHECK: alloca i{{[0-9]+}}, 1224 // CHECK: alloca i{{[0-9]+}}, 1225 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 1226 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], 1227 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_INT_TY]], 1228 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca i{{[0-9]+}}, 1229 1230 // Reduction list for runtime. 1231 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], 1232 1233 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1234 1235 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 1236 // CHECK: [[VAR1_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 1237 // CHECK: [[T_VAR1_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 1238 1239 // For + reduction operation initial value of private variable is 0. 1240 // CHECK: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[T_VAR_PRIV]], 1241 1242 // For & reduction operation initial value of private variable is ones in all bits. 1243 // CHECK: [[VAR_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 1244 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR_PRIV]]) 1245 1246 // For && reduction operation initial value of private variable is 1.0. 1247 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR1_PRIV]]) 1248 1249 // For min reduction operation initial value of private variable is largest repesentable value. 1250 // CHECK: store i{{[0-9]+}} 2147483647, i{{[0-9]+}}* [[T_VAR1_PRIV]], 1251 1252 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] 1253 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] 1254 // CHECK: call void @__kmpc_for_static_init_4( 1255 // Skip checks for internal operations. 1256 // CHECK: call void @__kmpc_for_static_fini( 1257 1258 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]}; 1259 1260 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0 1261 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR_PRIV]] to i8* 1262 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], 1263 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1 1264 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR_PRIV]] to i8* 1265 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], 1266 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2 1267 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_PRIV]] to i8* 1268 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], 1269 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3 1270 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR1_PRIV]] to i8* 1271 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], 1272 1273 // res = __kmpc_reduce_nowait(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>); 1274 1275 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* 1276 // 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]]) 1277 1278 // switch(res) 1279 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ 1280 // CHECK: i32 1, label %[[CASE1:.+]] 1281 // CHECK: i32 2, label %[[CASE2:.+]] 1282 // CHECK: ] 1283 1284 // case 1: 1285 // t_var += t_var_reduction; 1286 // CHECK: [[T_VAR_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_REF]], 1287 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]], 1288 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_VAL]], [[T_VAR_PRIV_VAL]] 1289 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_REF]], 1290 1291 // var = var.operator &(var_reduction); 1292 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 1293 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* 1294 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 1295 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 1296 1297 // var1 = var1.operator &&(var1_reduction); 1298 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) 1299 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1300 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 1301 // CHECK: [[TRUE]] 1302 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) 1303 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1304 // CHECK: br label %[[END2]] 1305 // CHECK: [[END2]] 1306 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 1307 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 1308 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 1309 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* 1310 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 1311 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 1312 1313 // t_var1 = min(t_var1, t_var1_reduction); 1314 // CHECK: [[T_VAR1_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_REF]], 1315 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]], 1316 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_VAL]], [[T_VAR1_PRIV_VAL]] 1317 // CHECK: br i1 [[CMP]] 1318 // CHECK: [[UP:%.+]] = phi i32 1319 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_REF]], 1320 1321 // __kmpc_end_reduce_nowait(<loc>, <gtid>, &<lock>); 1322 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) 1323 1324 // break; 1325 // CHECK: br label %[[RED_DONE]] 1326 1327 // case 2: 1328 // t_var += t_var_reduction; 1329 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]] 1330 // CHECK: atomicrmw add i32* [[T_VAR_REF]], i32 [[T_VAR_PRIV_VAL]] monotonic 1331 1332 // var = var.operator &(var_reduction); 1333 // CHECK: call void @__kmpc_critical( 1334 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* dereferenceable(4) [[VAR_PRIV]]) 1335 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* 1336 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 1337 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 1338 // CHECK: call void @__kmpc_end_critical( 1339 1340 // var1 = var1.operator &&(var1_reduction); 1341 // CHECK: call void @__kmpc_critical( 1342 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) 1343 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1344 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 1345 // CHECK: [[TRUE]] 1346 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) 1347 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1348 // CHECK: br label %[[END2]] 1349 // CHECK: [[END2]] 1350 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 1351 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 1352 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 1353 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* 1354 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 1355 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 1356 // CHECK: call void @__kmpc_end_critical( 1357 1358 // t_var1 = min(t_var1, t_var1_reduction); 1359 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]] 1360 // CHECK: atomicrmw min i32* [[T_VAR1_REF]], i32 [[T_VAR1_PRIV_VAL]] monotonic 1361 1362 // break; 1363 // CHECK: br label %[[RED_DONE]] 1364 // CHECK: [[RED_DONE]] 1365 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) 1366 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* 1367 // CHECK: ret void 1368 1369 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) { 1370 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); 1371 // ... 1372 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1], 1373 // *(Type<n>-1*)rhs[<n>-1]); 1374 // } 1375 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) 1376 // t_var_lhs = (i{{[0-9]+}}*)lhs[0]; 1377 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0 1378 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], 1379 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to i{{[0-9]+}}* 1380 // t_var_rhs = (i{{[0-9]+}}*)rhs[0]; 1381 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0 1382 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], 1383 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to i{{[0-9]+}}* 1384 1385 // var_lhs = (S<i{{[0-9]+}}>*)lhs[1]; 1386 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1 1387 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], 1388 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_INT_TY]]* 1389 // var_rhs = (S<i{{[0-9]+}}>*)rhs[1]; 1390 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1 1391 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], 1392 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_INT_TY]]* 1393 1394 // var1_lhs = (S<i{{[0-9]+}}>*)lhs[2]; 1395 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2 1396 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], 1397 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_INT_TY]]* 1398 // var1_rhs = (S<i{{[0-9]+}}>*)rhs[2]; 1399 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2 1400 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], 1401 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_INT_TY]]* 1402 1403 // t_var1_lhs = (i{{[0-9]+}}*)lhs[3]; 1404 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3 1405 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], 1406 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to i{{[0-9]+}}* 1407 // t_var1_rhs = (i{{[0-9]+}}*)rhs[3]; 1408 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3 1409 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], 1410 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to i{{[0-9]+}}* 1411 1412 // t_var_lhs += t_var_rhs; 1413 // CHECK: [[T_VAR_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_LHS]], 1414 // CHECK: [[T_VAR_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_RHS]], 1415 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_LHS_VAL]], [[T_VAR_RHS_VAL]] 1416 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_LHS]], 1417 1418 // var_lhs = var_lhs.operator &(var_rhs); 1419 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_LHS]], [[S_INT_TY]]* dereferenceable(4) [[VAR_RHS]]) 1420 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_LHS]] to i8* 1421 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* 1422 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 1423 1424 // var1_lhs = var1_lhs.operator &&(var1_rhs); 1425 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_LHS]]) 1426 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1427 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]] 1428 // CHECK: [[TRUE]] 1429 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_RHS]]) 1430 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 1431 // CHECK: br label %[[END2]] 1432 // CHECK: [[END2]] 1433 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ] 1434 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32 1435 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]]) 1436 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_LHS]] to i8* 1437 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* 1438 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[BC1]], i8* align 4 [[BC2]], i64 4, i1 false) 1439 1440 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs); 1441 // CHECK: [[T_VAR1_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_LHS]], 1442 // CHECK: [[T_VAR1_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_RHS]], 1443 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_LHS_VAL]], [[T_VAR1_RHS_VAL]] 1444 // CHECK: br i1 [[CMP]] 1445 // CHECK: [[UP:%.+]] = phi i32 1446 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_LHS]], 1447 // CHECK: ret void 1448 1449 // 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) %{{.*}}) 1450 1451 // CHECK: [[ARR_ORIG_ADDR:%.+]] = alloca [42 x [[S_INT_TY]]]*, 1452 // CHECK: [[ARR_PRIV:%.+]] = alloca [40 x [[S_INT_TY]]], 1453 1454 // Reduction list for runtime. 1455 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*], 1456 1457 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 1458 1459 // CHECK: [[ARR_ORIG:%.+]] = load [42 x [[S_INT_TY]]]*, [42 x [[S_INT_TY]]]** [[ARR_ORIG_ADDR]], 1460 // CHECK: [[LOW:%.+]] = getelementptr inbounds [42 x [[S_INT_TY]]], [42 x [[S_INT_TY]]]* [[ARR_ORIG]], i64 0, i64 1 1461 // CHECK: [[ORIG_START:%.+]] = bitcast [42 x [[S_INT_TY]]]* [[ARR_ORIG]] to [[S_INT_TY]]* 1462 // CHECK: [[START:%.+]] = ptrtoint [[S_INT_TY]]* [[ORIG_START]] to i64 1463 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_INT_TY]]* [[LOW]] to i64 1464 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]] 1465 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64) 1466 // CHECK: [[ARR_PRIV_PTR:%.+]] = bitcast [40 x [[S_INT_TY]]]* [[ARR_PRIV]] to [[S_INT_TY]]* 1467 // CHECK: [[PSEUDO_ARR_PRIV:%.+]] = getelementptr [[S_INT_TY]], [[S_INT_TY]]* [[ARR_PRIV_PTR]], i64 [[OFFSET]] 1468 // CHECK: [[ARR_PRIV:%.+]] = bitcast [[S_INT_TY]]* [[PSEUDO_ARR_PRIV]] to [42 x [[S_INT_TY]]]* 1469 1470 // CHECK: ret void 1471 1472 #endif 1473 1474