1 // RUN: %clang_cc1 -DLAMBDA -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-target | FileCheck %s --check-prefix LAMBDA --check-prefix LAMBDA-64 2 // RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 3 // RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-target | FileCheck %s --check-prefix LAMBDA --check-prefix LAMBDA-64 4 // RUN: %clang_cc1 -DLAMBDA -verify -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-target | FileCheck %s --check-prefix LAMBDA --check-prefix LAMBDA-32 5 // RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 6 // RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-target | FileCheck %s --check-prefix LAMBDA --check-prefix LAMBDA-32 7 8 // RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 9 // RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-target | FileCheck --check-prefix SIMD-ONLY0 %s 10 // RUN: %clang_cc1 -DLAMBDA -verify -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-target | FileCheck --check-prefix SIMD-ONLY0 %s 11 // RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 12 // RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-target | FileCheck --check-prefix SIMD-ONLY0 %s 13 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}} 14 15 // RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-target | FileCheck %s --check-prefix CHECK --check-prefix CHECK-64 16 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 17 // RUN: %clang_cc1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-target | FileCheck %s --check-prefix CHECK --check-prefix CHECK-64 18 // RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-target | FileCheck %s --check-prefix CHECK --check-prefix CHECK-32 19 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 20 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-target | FileCheck %s --check-prefix CHECK --check-prefix CHECK-32 21 22 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-target | FileCheck --check-prefix SIMD-ONLY1 %s 23 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 24 // RUN: %clang_cc1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-target | FileCheck --check-prefix SIMD-ONLY1 %s 25 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-target | FileCheck --check-prefix SIMD-ONLY1 %s 26 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 27 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-target | FileCheck --check-prefix SIMD-ONLY1 %s 28 // SIMD-ONLY1-NOT: {{__kmpc|__tgt}} 29 // expected-no-diagnostics 30 #ifndef HEADER 31 #define HEADER 32 33 template <class T> 34 struct S { 35 T f; 36 S(T a) : f(a) {} 37 S() : f() {} 38 operator T() { return T(); } 39 ~S() {} 40 }; 41 42 // CHECK: [[S_FLOAT_TY:%.+]] = type { float } 43 // CHECK: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} } 44 template <typename T> 45 T tmain() { 46 S<T> test; 47 T t_var = T(); 48 T vec[] = {1, 2}; 49 S<T> s_arr[] = {1, 2}; 50 S<T> &var = test; 51 #pragma omp target 52 #pragma omp teams 53 #pragma omp distribute parallel for simd firstprivate(t_var, vec, s_arr, s_arr, var, var) 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 int main() { 62 static int svar; 63 volatile double g; 64 volatile double &g1 = g; 65 66 #ifdef LAMBDA 67 // LAMBDA-LABEL: @main 68 // LAMBDA: call{{.*}} void [[OUTER_LAMBDA:@.+]]( 69 [&]() { 70 static float sfvar; 71 // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]]( 72 // LAMBDA: call i{{[0-9]+}} @__tgt_target_teams( 73 // LAMBDA: call void [[OFFLOADING_FUN:@.+]]( 74 75 // LAMBDA: define{{.+}} void [[OFFLOADING_FUN]]( 76 // LAMBDA: call {{.*}}void {{.+}} @__kmpc_fork_teams({{.+}}, i32 4, {{.+}}* [[OMP_OUTLINED:@.+]] to {{.+}}) 77 #pragma omp target 78 #pragma omp teams 79 #pragma omp distribute parallel for simd firstprivate(g, g1, svar, sfvar) 80 for (int i = 0; i < 2; ++i) { 81 // LAMBDA: define{{.*}} internal{{.*}} void [[OMP_OUTLINED]](i32* noalias %{{.+}}, i32* noalias %{{.+}}, double* {{.+}} [[G_IN:%.+]], double*{{.+}} [[G1_IN:%.+]], i{{[0-9]+}}*{{.+}} [[SVAR_IN:%.+]], float*{{.+}} [[SFVAR_IN:%.+]]) 82 83 // addr alloca's 84 // LAMBDA: [[G_ADDR:%.+]] = alloca double*, 85 // LAMBDA: [[G1_ADDR:%.+]] = alloca double*, 86 // LAMBDA: [[SVAR_ADDR:%.+]] = alloca i{{[0-9]+}}*, 87 // LAMBDA: [[SFVAR_ADDR:%.+]] = alloca float*, 88 // LAMBDA: [[G1_REF:%.+]] = alloca double*, 89 90 // private alloca's 91 // LAMBDA: [[G_PRIV:%.+]] = alloca double, 92 // LAMBDA: [[G1_PRIV:%.+]] = alloca double, 93 // LAMBDA: [[TMP_PRIV:%.+]] = alloca double*, 94 // LAMBDA: [[SVAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 95 // LAMBDA: [[SFVAR_PRIV:%.+]] = alloca float, 96 97 // transfer input parameters into addr alloca's 98 // LAMBDA-DAG: store {{.+}} [[G_IN]], {{.+}} [[G_ADDR]], 99 // LAMBDA-DAG: store {{.+}} [[G1_IN]], {{.+}} [[G1_ADDR]], 100 // LAMBDA-DAG: store {{.+}} [[SVAR_IN]], {{.+}} [[SVAR_ADDR]], 101 // LAMBDA-DAG: store {{.+}} [[SFVAR_IN]], {{.+}} [[SFVAR_ADDR]], 102 103 // init private alloca's with addr alloca's 104 // g 105 // LAMBDA-DAG: [[G_CONV:%.+]] = load {{.+}}*, {{.+}}** [[G_ADDR]] 106 // LAMBDA-DAG: [[G_ADDR_VAL:%.+]] = load {{.+}}, {{.+}}* [[G_CONV]], 107 // LAMBDA-DAG: store {{.+}} [[G_ADDR_VAL]], {{.+}}* [[G_PRIV]], 108 109 // g1 110 // LAMBDA-DAG: [[TMP_REF:%.+]] = load {{.+}}*, {{.+}}** [[G1_REF]], 111 // LAMBDA-DAG: [[TMP_VAL:%.+]] = load {{.+}}, {{.+}}* [[TMP_REF]], 112 // LAMBDA-DAG: store {{.+}} [[TMP_VAL]], {{.+}}* [[G1_PRIV]] 113 // LAMBDA-DAG: store {{.+}}* [[G1_PRIV]], {{.+}}** [[TMP_PRIV]], 114 115 // svar 116 // LAMBDA-DAG: [[SVAR_REF:%.+]] = load {{.+}}*, {{.+}}** [[SVAR_ADDR]], 117 // LAMBDA-DAG: [[SVAR_VAL:%.+]] = load {{.+}}, {{.+}}* [[SVAR_REF]], 118 // LAMBDA-DAG: store {{.+}} [[SVAR_VAL]], {{.+}}* [[SVAR_PRIV]], 119 120 // sfvar 121 // LAMBDA-DAG: [[SFVAR_REF:%.+]] = load {{.+}}*, {{.+}}** [[SFVAR_ADDR]], 122 // LAMBDA-DAG: [[SFVAR_VAL:%.+]] = load {{.+}}, {{.+}}* [[SFVAR_REF]], 123 // LAMBDA-DAG: store {{.+}} [[SFVAR_VAL]], {{.+}}* [[SFVAR_PRIV]], 124 125 // LAMBDA: call {{.*}}void @__kmpc_for_static_init_4( 126 // pass firstprivate parameters to parallel outlined function 127 // g 128 // LAMBDA-64-DAG: [[G_PRIV_VAL:%.+]] = load {{.+}}, {{.+}}* [[G_PRIV]], 129 // LAMBDA-64: [[G_CAST_CONV:%.+]] = bitcast {{.+}}* [[G_CAST:%.+]] to 130 // LAMBDA-64-DAG: store {{.+}} [[G_PRIV_VAL]], {{.+}}* [[G_CAST_CONV]], 131 // LAMBDA-64-DAG: [[G_PAR:%.+]] = load {{.+}}, {{.+}}* [[G_CAST]], 132 133 // g1 134 // LAMBDA-DAG: [[TMP_PRIV_VAL:%.+]] = load {{.+}}, {{.+}}* [[TMP_PRIV]], 135 // LAMBDA-DAG: [[G1_PRIV_VAL:%.+]] = load {{.+}}, {{.+}}* [[TMP_PRIV_VAL]], 136 // LAMBDA: [[G1_CAST_CONV:%.+]] = bitcast {{.+}}* [[G1_CAST:%.+]] to 137 // LAMBDA-DAG: store {{.+}} [[G1_PRIV_VAL]], {{.+}}* [[G1_CAST_CONV]], 138 // LAMBDA-DAG: [[G1_PAR:%.+]] = load {{.+}}, {{.+}}* [[G1_CAST]], 139 140 // svar 141 // LAMBDA: [[SVAR_VAL:%.+]] = load {{.+}}, {{.+}}* [[SVAR_PRIV]], 142 // LAMBDA-64-DAG: [[SVAR_CAST_CONV:%.+]] = bitcast {{.+}}* [[SVAR_CAST:%.+]] to 143 // LAMBDA-64-DAG: store {{.+}} [[SVAR_VAL]], {{.+}}* [[SVAR_CAST_CONV]], 144 // LAMBDA-32-DAG: store {{.+}} [[SVAR_VAL]], {{.+}}* [[SVAR_CAST:%.+]], 145 // LAMBDA-DAG: [[SVAR_PAR:%.+]] = load {{.+}}, {{.+}}* [[SVAR_CAST]], 146 147 // sfvar 148 // LAMBDA: [[SFVAR_VAL:%.+]] = load {{.+}}, {{.+}}* [[SFVAR_PRIV]], 149 // LAMBDA-DAG: [[SFVAR_CAST_CONV:%.+]] = bitcast {{.+}}* [[SFVAR_CAST:%.+]] to 150 // LAMBDA-DAG: store {{.+}} [[SFVAR_VAL]], {{.+}}* [[SFVAR_CAST_CONV]], 151 // LAMBDA-DAG: [[SFVAR_PAR:%.+]] = load {{.+}}, {{.+}}* [[SFVAR_CAST]], 152 153 // LAMBDA-64: call{{.+}} @__kmpc_fork_call({{.+}}, {{.+}}, {{.+}}[[OMP_PARFOR_OUTLINED:@.+]] to void ({{.+}})*), {{.+}}, {{.+}}, {{.+}} [[G_PAR]], {{.+}} [[G1_PAR]], {{.+}} [[SVAR_PAR]], {{.+}} [[SFVAR_PAR]]) 154 // LAMBDA-32: call{{.+}} @__kmpc_fork_call({{.+}}, {{.+}}, {{.+}}[[OMP_PARFOR_OUTLINED:@.+]] to void ({{.+}})*), {{.+}}, {{.+}}, {{.+}} [[G_PRIV]], {{.+}} [[G1_PAR]], {{.+}} [[SVAR_PAR]], {{.+}} [[SFVAR_PAR]]) 155 // LAMBDA: call {{.*}}void @__kmpc_for_static_fini( 156 // LAMBDA: ret void 157 158 159 // LAMBDA-64: define{{.+}} void [[OMP_PARFOR_OUTLINED]](i32* noalias %{{.+}}, i32* noalias %{{.+}}, {{.+}}, {{.+}}, i{{[0-9]+}} [[G_IN:%.+]], i{{[0-9]+}} [[G1_IN:%.+]], i{{[0-9]+}} [[SVAR_IN:%.+]], i{{[0-9]+}} [[SFVAR_IN:%.+]]) 160 // LAMBDA-32: define{{.+}} void [[OMP_PARFOR_OUTLINED]](i32* noalias %{{.+}}, i32* noalias %{{.+}}, {{.+}}, {{.+}}, double* {{.+}} [[G_IN:%.+]], i{{[0-9]+}} [[G1_IN:%.+]], i{{[0-9]+}} [[SVAR_IN:%.+]], i{{[0-9]+}} [[SFVAR_IN:%.+]]) 161 // skip initial params 162 // LAMBDA: {{.+}} = alloca{{.+}}, 163 // LAMBDA: {{.+}} = alloca{{.+}}, 164 // LAMBDA: {{.+}} = alloca{{.+}}, 165 // LAMBDA: {{.+}} = alloca{{.+}}, 166 167 // addr alloca's 168 // LAMBDA-64: [[G_ADDR:%.+]] = alloca i{{[0-9]+}}, 169 // LAMBDA-32: [[G_ADDR:%.+]] = alloca double*, 170 // LAMBDA: [[G1_ADDR:%.+]] = alloca i{{[0-9]+}}, 171 // LAMBDA: [[SVAR_ADDR:%.+]] = alloca i{{[0-9]+}}, 172 // LAMBDA: [[SFVAR_ADDR:%.+]] = alloca i{{[0-9]+}}, 173 // LAMBDA: [[G1_REF:%.+]] = alloca double*, 174 175 // private alloca's (only for 32-bit) 176 // LAMBDA-32: [[G_PRIV:%.+]] = alloca double, 177 178 // transfer input parameters into addr alloca's 179 // LAMBDA-DAG: store {{.+}} [[G_IN]], {{.+}} [[G_ADDR]], 180 // LAMBDA-DAG: store {{.+}} [[G1_IN]], {{.+}} [[G1_ADDR]], 181 // LAMBDA-DAG: store {{.+}} [[SVAR_IN]], {{.+}} [[SVAR_ADDR]], 182 // LAMBDA-DAG: store {{.+}} [[SFVAR_IN]], {{.+}} [[SFVAR_ADDR]], 183 184 // prepare parameters for lambda 185 // g 186 // LAMBDA-64-DAG: [[G_CONV:%.+]] = bitcast {{.+}}* [[G_ADDR]] to 187 // LAMBDA-32-DAG: [[G_ADDR_REF:%.+]] = load {{.+}}*, {{.+}}** [[G_ADDR]] 188 // LAMBDA-32-DAG: [[G_ADDR_VAL:%.+]] = load {{.+}}, {{.+}}* [[G_ADDR_REF]], 189 // LAMBDA-32-DAG: store {{.+}} [[G_ADDR_VAL]], {{.+}}* [[G_PRIV]], 190 191 // g1 192 // LAMBDA-DAG: [[G1_CONV:%.+]] = bitcast {{.+}}* [[G1_ADDR]] to 193 // LAMBDA-DAG: store {{.+}}* [[G1_CONV]], {{.+}}* [[G1_REF]], 194 195 // svar 196 // LAMBDA-64-DAG: [[SVAR_CONV:%.+]] = bitcast {{.+}}* [[SVAR_ADDR]] to 197 198 // sfvar 199 // LAMBDA-DAG: [[SFVAR_CONV:%.+]] = bitcast {{.+}}* [[SFVAR_ADDR]] to 200 201 // LAMBDA: call {{.*}}void @__kmpc_for_static_init_4( 202 g = 1; 203 g1 = 1; 204 svar = 3; 205 sfvar = 4.0; 206 // LAMBDA-64: store double 1.0{{.+}}, double* [[G_CONV]], 207 // LAMBDA-32: store double 1.0{{.+}}, double* [[G_PRIV]], 208 // LAMBDA: [[G1_REF_REF:%.+]] = load {{.+}}*, {{.+}}** [[G1_REF]], 209 // LAMBDA: store {{.+}} 1.0{{.+}}, {{.+}}* [[G1_REF_REF]], 210 // LAMBDA-64: store {{.+}} 3, {{.+}}* [[SVAR_CONV]], 211 // LAMBDA-32: store {{.+}} 3, {{.+}}* [[SVAR_ADDR]], 212 // LAMBDA: store {{.+}} 4.0{{.+}}, {{.+}}* [[SFVAR_CONV]], 213 214 // pass params to inner lambda 215 // LAMBDA: [[G_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 216 // LAMBDA-64: store double* [[G_CONV]], double** [[G_PRIVATE_ADDR_REF]], 217 // LAMBDA-32: store double* [[G_PRIV]], double** [[G_PRIVATE_ADDR_REF]], 218 // LAMBDA: [[G1_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 1 219 // LAMBDA: [[G1_REF_REF:%.+]] = load double*, double** [[G1_REF]], 220 // LAMBDA: store double* [[G1_REF_REF]], double** [[G1_PRIVATE_ADDR_REF]], 221 // LAMBDA: [[SVAR_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 2 222 // LAMBDA-64: store i{{[0-9]+}}* [[SVAR_CONV]], i{{[0-9]+}}** [[SVAR_PRIVATE_ADDR_REF]] 223 // LAMBDA-32: store i{{[0-9]+}}* [[SVAR_ADDR]], i{{[0-9]+}}** [[SVAR_PRIVATE_ADDR_REF]] 224 // LAMBDA: [[SFVAR_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 3 225 // LAMBDA: store float* [[SFVAR_CONV]], float** [[SFVAR_PRIVATE_ADDR_REF]] 226 // LAMBDA: call{{.*}} void [[INNER_LAMBDA:@.+]](%{{.+}}* [[ARG]]) 227 // LAMBDA: call {{.*}}void @__kmpc_for_static_fini( 228 // LAMBDA: ret void 229 [&]() { 230 // LAMBDA: define {{.+}} void [[INNER_LAMBDA]](%{{.+}}* [[ARG_PTR:%.+]]) 231 // LAMBDA: store %{{.+}}* [[ARG_PTR]], %{{.+}}** [[ARG_PTR_REF:%.+]], 232 g = 2; 233 g1 = 2; 234 svar = 4; 235 sfvar = 8.0; 236 // LAMBDA: [[ARG_PTR:%.+]] = load %{{.+}}*, %{{.+}}** [[ARG_PTR_REF]] 237 // LAMBDA: [[G_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 238 // LAMBDA: [[G_REF:%.+]] = load double*, double** [[G_PTR_REF]] 239 // LAMBDA: store double 2.0{{.+}}, double* [[G_REF]] 240 241 // LAMBDA: [[TMP_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 1 242 // LAMBDA: [[G1_REF:%.+]] = load double*, double** [[TMP_PTR_REF]] 243 // LAMBDA: store double 2.0{{.+}}, double* [[G1_REF]], 244 // LAMBDA: [[SVAR_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 2 245 // LAMBDA: [[SVAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[SVAR_PTR_REF]] 246 // LAMBDA: store i{{[0-9]+}} 4, i{{[0-9]+}}* [[SVAR_REF]] 247 // LAMBDA: [[SFVAR_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 3 248 // LAMBDA: [[SFVAR_REF:%.+]] = load float*, float** [[SFVAR_PTR_REF]] 249 // LAMBDA: store float 8.0{{.+}}, float* [[SFVAR_REF]] 250 }(); 251 } 252 }(); 253 return 0; 254 #else 255 S<float> test; 256 int t_var = 0; 257 int vec[] = {1, 2}; 258 S<float> s_arr[] = {1, 2}; 259 S<float> &var = test; 260 261 #pragma omp target 262 #pragma omp teams 263 #pragma omp distribute parallel for simd firstprivate(t_var, vec, s_arr, s_arr, var, var, svar) 264 for (int i = 0; i < 2; ++i) { 265 vec[i] = t_var; 266 s_arr[i] = var; 267 } 268 return tmain<int>(); 269 #endif 270 } 271 272 // CHECK-LABEL: define{{.*}} i{{[0-9]+}} @main() 273 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]], 274 // CHECK: call {{.*}} [[S_FLOAT_TY_DEF_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]]) 275 // CHECK: call i{{[0-9]+}} @__tgt_target_teams( 276 // CHECK: call void [[OFFLOAD_FUN_0:@.+]]( 277 // CHECK: call {{.*}} [[S_FLOAT_TY_DEF_DESTR:@.+]]([[S_FLOAT_TY]]* [[TEST]]) 278 279 // CHECK: define{{.+}} [[OFFLOAD_FUN_0]](i{{[0-9]+}} [[T_VAR_IN:%.+]], [2 x i{{[0-9]+}}]* {{.+}} [[VEC_IN:%.+]], [2 x [[S_FLOAT_TY]]]* {{.+}} [[S_ARR_IN:%.+]], [[S_FLOAT_TY]]* {{.+}} [[VAR_IN:%.+]], i{{[0-9]+}} [[SVAR_IN:%.+]]) 280 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_teams(%{{.+}}* @{{.+}}, i{{[0-9]+}} 5, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, i{{[0-9]+}}*, [2 x i{{[0-9]+}}]*, [2 x [[S_FLOAT_TY]]]*, [[S_FLOAT_TY]]*, i{{[0-9]+}}*)* [[OMP_OUTLINED_0:@.+]] to void 281 // CHECK: ret 282 283 // CHECK: define internal void [[OMP_OUTLINED_0]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i{{[0-9]+}}*{{.+}} [[T_VAR_IN:%.+]], [2 x i{{[0-9]+}}]* {{.+}} [[VEC_IN:%.+]], [2 x [[S_FLOAT_TY]]]* {{.+}} [[S_ARR_IN:%.+]], [[S_FLOAT_TY]]* {{.+}} [[VAR_IN:%.+]], i{{[0-9]+}}*{{.+}} [[SVAR_IN:%.+]]) 284 285 // CHECK: alloca i{{[0-9]+}}*, 286 // CHECK: alloca i{{[0-9]+}}*, 287 // addr alloca's 288 // CHECK: [[T_VAR_ADDR:%.+]] = alloca i{{[0-9]+}}*, 289 // CHECK: [[VEC_ADDR:%.+]] = alloca [2 x i{{[0-9]+}}]*, 290 // CHECK: [[S_ARR_ADDR:%.+]] = alloca [2 x [[S_FLOAT_TY]]]*, 291 // CHECK: [[VAR_ADDR:%.+]] = alloca [[S_FLOAT_TY]]*, 292 // CHECK: [[SVAR_ADDR:%.+]] = alloca i{{[0-9]+}}*, 293 // CHECK: [[TMP:%.+]] = alloca [[S_FLOAT_TY]]*, 294 295 // skip loop alloca's 296 // CHECK: [[OMP_IV:.omp.iv+]] = alloca i{{[0-9]+}}, 297 // CHECK: [[OMP_LB:.omp.comb.lb+]] = alloca i{{[0-9]+}}, 298 // CHECK: [[OMP_UB:.omp.comb.ub+]] = alloca i{{[0-9]+}}, 299 // CHECK: [[OMP_ST:.omp.stride+]] = alloca i{{[0-9]+}}, 300 // CHECK: [[OMP_IS_LAST:.omp.is_last+]] = alloca i{{[0-9]+}}, 301 302 // private alloca's 303 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 304 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}], 305 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]], 306 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 307 // CHECK: [[TMP_PRIV:%.+]] = alloca [[S_FLOAT_TY]]*, 308 // CHECK: [[SVAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 309 310 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_REF:%.+]] 311 312 // init addr alloca's with input values 313 // CHECK-DAG: store {{.+}} [[T_VAR_IN]], {{.+}}* [[T_VAR_ADDR]], 314 // CHECK-DAG: store {{.+}} [[VEC_IN]], {{.+}} [[VEC_ADDR]], 315 // CHECK-DAG: store {{.+}} [[S_ARR_IN]], {{.+}} [[S_ARR_ADDR]], 316 // CHECK-DAG: store {{.+}} [[VAR_IN]], {{.+}} [[VAR_ADDR]], 317 // CHECK-DAG: store {{.+}} [[SVAR_IN]], {{.+}} [[SVAR_ADDR]], 318 319 // init private alloca's with addr alloca's 320 // t-var 321 // CHECK-DAG: [[T_VAR_ADDR_REF:%.+]] = load {{.+}}*, {{.+}}** [[T_VAR_ADDR]], 322 // CHECK-DAG: [[T_VAR:%.+]] = load {{.+}}, {{.+}}* [[T_VAR_ADDR_REF]], 323 // CHECK-DAG: store {{.+}} [[T_VAR]], {{.+}} [[T_VAR_PRIV]], 324 325 // vec 326 // CHECK-DAG: [[VEC_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[VEC_ADDR]], 327 // CHECK-DAG: [[VEC_PRIV_BCAST:%.+]] = bitcast {{.+}} [[VEC_PRIV]] to 328 // CHECK-DAG: [[VEC_ADDR_BCAST:%.+]] = bitcast {{.+}} [[VEC_ADDR_VAL]] to 329 // CHECK-DAG: call void @llvm.memcpy{{.+}}({{.+}}* align {{[0-9]+}} [[VEC_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[VEC_ADDR_BCAST]], 330 331 // s_arr 332 // CHECK-DAG: [[S_ARR_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[S_ARR_ADDR]], 333 // CHECK-DAG: [[S_ARR_BGN:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_PRIV]], 334 // CHECK-DAG: [[S_ARR_ADDR_BCAST:%.+]] = bitcast {{.+}}* [[S_ARR_ADDR_VAL]] to 335 // CHECK-DAG: [[S_ARR_BGN_GEP:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_BGN]], 336 // CHECK-DAG: [[S_ARR_EMPTY:%.+]] = icmp {{.+}} [[S_ARR_BGN]], [[S_ARR_BGN_GEP]] 337 // CHECK-DAG: br {{.+}} [[S_ARR_EMPTY]], label %[[CPY_DONE:.+]], label %[[CPY_BODY:.+]] 338 // CHECK-DAG: [[CPY_BODY]]: 339 // CHECK-DAG: call void @llvm.memcpy{{.+}}( 340 // CHECK-DAG: [[CPY_DONE]]: 341 342 // var 343 // CHECK-DAG: [[TMP_REF:%.+]] = load {{.+}}*, {{.+}}* [[TMP]], 344 // CHECK-DAG: [[VAR_PRIV_BCAST:%.+]] = bitcast {{.+}}* [[VAR_PRIV]] to 345 // CHECK-DAG: [[TMP_REF_BCAST:%.+]] = bitcast {{.+}}* [[TMP_REF]] to 346 // CHECK-DAG: call void @llvm.memcpy.{{.+}}({{.+}}* align {{[0-9]+}} [[VAR_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[TMP_REF_BCAST]], 347 // CHECK-DAG: store {{.+}}* [[VAR_PRIV]], {{.+}}** [[TMP_PRIV]], 348 349 // svar 350 // CHECK-DAG: [[SVAR_CONV_REF:%.+]] = load {{.+}}*, {{.+}}** [[SVAR_ADDR]], 351 // CHECK-DAG: [[SVAR_CONV_VAL:%.+]] = load {{.+}}, {{.+}}* [[SVAR_CONV_REF]], 352 // CHECK-DAG: store {{.+}} [[SVAR_CONV_VAL]], {{.+}}* [[SVAR_PRIV]], 353 354 // CHECK: call void @__kmpc_for_static_init_4( 355 // pass private alloca's to fork 356 // CHECK-DAG: [[T_VAR_PRIV_VAL:%.+]] = load {{.+}}, {{.+}}* [[T_VAR_PRIV]], 357 // not dag to distinguish with S_VAR_CAST 358 // CHECK-64: [[T_VAR_CAST_CONV:%.+]] = bitcast {{.+}}* [[T_VAR_CAST:%.+]] to 359 // CHECK-64-DAG: store {{.+}} [[T_VAR_PRIV_VAL]], {{.+}} [[T_VAR_CAST_CONV]], 360 // CHECK-32: store {{.+}} [[T_VAR_PRIV_VAL]], {{.+}} [[T_VAR_CAST:%.+]], 361 // CHECK-DAG: [[T_VAR_CAST_VAL:%.+]] = load {{.+}}, {{.+}}* [[T_VAR_CAST]], 362 // CHECK-DAG: [[TMP_PRIV_VAL:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[TMP_PRIV]], 363 // CHECK-DAG: [[SVAR_PRIV_VAL:%.+]] = load {{.+}}, {{.+}}* [[SVAR_PRIV]], 364 // CHECK-64-DAG: [[SVAR_CAST_CONV:%.+]] = bitcast {{.+}}* [[SVAR_CAST:%.+]] to 365 // CHECK-64-DAG: store {{.+}} [[SVAR_PRIV_VAL]], {{.+}}* [[SVAR_CAST_CONV]], 366 // CHECK-32-DAG: store {{.+}} [[SVAR_PRIV_VAL]], {{.+}}* [[SVAR_CAST:%.+]], 367 // CHECK-DAG: [[SVAR_CAST_VAL:%.+]] = load {{.+}}, {{.+}}* [[SVAR_CAST]], 368 // CHECK: call{{.+}} @__kmpc_fork_call({{.+}}, {{.+}}, {{.+}}[[OMP_PARFOR_OUTLINED_0:@.+]] to void ({{.+}})*), {{.+}}, {{.+}}, [2 x i{{[0-9]+}}]* [[VEC_PRIV]], i{{[0-9]+}} [[T_VAR_CAST_VAL]], [2 x [[S_FLOAT_TY]]]* [[S_ARR_PRIV]], [[S_FLOAT_TY]]* [[TMP_PRIV_VAL]], i{{[0-9]+}} [[SVAR_CAST_VAL]]) 369 // CHECK: call void @__kmpc_for_static_fini( 370 371 // call destructors: var.. 372 // CHECK-DAG: call {{.+}} [[S_FLOAT_TY_DEF_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 373 374 // ..and s_arr 375 // CHECK: {{.+}}: 376 // CHECK: [[S_ARR_EL_PAST:%.+]] = phi [[S_FLOAT_TY]]* 377 // CHECK: [[S_ARR_PRIV_ITEM:%.+]] = getelementptr {{.+}}, {{.+}} [[S_ARR_EL_PAST]], 378 // CHECK: call {{.*}} [[S_FLOAT_TY_DEF_DESTR]]([[S_FLOAT_TY]]* [[S_ARR_PRIV_ITEM]]) 379 380 // CHECK: ret void 381 382 // By OpenMP specifications, 'firstprivate' applies to both distribute and parallel for. 383 // However, the support for 'firstprivate' of 'parallel' is only used when 'parallel' 384 // is found alone. Therefore we only have one 'firstprivate' support for 'parallel for' 385 // in combination 386 // CHECK: define internal void [[OMP_PARFOR_OUTLINED_0]]({{.+}}, {{.+}}, {{.+}}, {{.+}}, [2 x i{{[0-9]+}}]* {{.+}} [[VEC_IN:%.+]], i{{[0-9]+}} [[T_VAR_IN:%.+]], [2 x [[S_FLOAT_TY]]]* {{.+}} [[S_ARR_IN:%.+]], [[S_FLOAT_TY]]* {{.+}} [[VAR_IN:%.+]], i{{[0-9]+}} [[SVAR_IN:%.+]]) 387 388 // addr alloca's 389 // CHECK: [[VEC_ADDR:%.+]] = alloca [2 x i{{[0-9]+}}]*, 390 // CHECK: [[T_VAR_ADDR:%.+]] = alloca i{{[0-9]+}}, 391 // CHECK: [[S_ARR_ADDR:%.+]] = alloca [2 x [[S_FLOAT_TY]]]*, 392 // CHECK: [[VAR_ADDR:%.+]] = alloca [[S_FLOAT_TY]]*, 393 // CHECK: [[SVAR_ADDR:%.+]] = alloca i{{[0-9]+}}, 394 395 // skip loop alloca's 396 // CHECK: [[OMP_IV:.omp.iv+]] = alloca i{{[0-9]+}}, 397 // CHECK: [[OMP_LB:.omp.lb+]] = alloca i{{[0-9]+}}, 398 // CHECK: [[OMP_UB:.omp.ub+]] = alloca i{{[0-9]+}}, 399 // CHECK: [[OMP_ST:.omp.stride+]] = alloca i{{[0-9]+}}, 400 // CHECK: [[OMP_IS_LAST:.omp.is_last+]] = alloca i{{[0-9]+}}, 401 402 // private alloca's 403 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}], 404 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]], 405 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 406 // CHECK: [[TMP_PRIV:%.+]] = alloca [[S_FLOAT_TY]]*, 407 408 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_REF:%.+]] 409 410 // init addr alloca's with input values 411 // CHECK-DAG: store {{.+}} [[VEC_IN]], {{.+}} [[VEC_ADDR]], 412 // CHECK-DAG: store {{.+}} [[T_VAR_IN]], {{.+}}* [[T_VAR_ADDR]], 413 // CHECK-DAG: store {{.+}} [[S_ARR_IN]], {{.+}} [[S_ARR_ADDR]], 414 // CHECK-DAG: store {{.+}} [[VAR_IN]], {{.+}} [[VAR_ADDR]], 415 // CHECK-DAG: store {{.+}} [[SVAR_IN]], {{.+}} [[SVAR_ADDR]], 416 417 // init private alloca's with addr alloca's 418 // vec 419 // CHECK-DAG: [[VEC_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[VEC_ADDR]], 420 // CHECK-DAG: [[VEC_PRIV_BCAST:%.+]] = bitcast {{.+}} [[VEC_PRIV]] to 421 // CHECK-DAG: [[VEC_ADDR_BCAST:%.+]] = bitcast {{.+}} [[VEC_ADDR_VAL]] to 422 // CHECK-DAG: call void @llvm.memcpy{{.+}}({{.+}}* align {{[0-9]+}} [[VEC_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[VEC_ADDR_BCAST]], 423 424 // s_arr 425 // CHECK-DAG: [[S_ARR_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[S_ARR_ADDR]], 426 // CHECK-DAG: [[S_ARR_BGN:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_PRIV]], 427 // CHECK-DAG: [[S_ARR_ADDR_BCAST:%.+]] = bitcast {{.+}}* [[S_ARR_ADDR_VAL]] to 428 // CHECK-DAG: [[S_ARR_BGN_GEP:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_BGN]], 429 // CHECK-DAG: [[S_ARR_EMPTY:%.+]] = icmp {{.+}} [[S_ARR_BGN]], [[S_ARR_BGN_GEP]] 430 // CHECK-DAG: br {{.+}} [[S_ARR_EMPTY]], label %[[CPY_DONE:.+]], label %[[CPY_BODY:.+]] 431 // CHECK-DAG: [[CPY_BODY]]: 432 // CHECK-DAG: call void @llvm.memcpy{{.+}}( 433 // CHECK-DAG: [[CPY_DONE]]: 434 435 // var 436 // CHECK-DAG: [[VAR_ADDR_REF:%.+]] = load {{.+}}*, {{.+}}* [[VAR_ADDR]], 437 // CHECK-DAG: [[VAR_PRIV_BCAST:%.+]] = bitcast {{.+}}* [[VAR_PRIV]] to 438 // CHECK-DAG: [[VAR_ADDR_BCAST:%.+]] = bitcast {{.+}}* [[VAR_ADDR_REF]] to 439 // CHECK-DAG: call void @llvm.memcpy.{{.+}}({{.+}}* align {{[0-9]+}} [[VAR_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[VAR_ADDR_BCAST]], 440 // CHECK-DAG: store {{.+}}* [[VAR_PRIV]], {{.+}}** [[TMP_PRIV]], 441 442 // CHECK: call void @__kmpc_for_static_init_4( 443 // CHECK: call void @__kmpc_for_static_fini( 444 445 // call destructors: var.. 446 // CHECK-DAG: call {{.+}} [[S_FLOAT_TY_DEF_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 447 448 // ..and s_arr 449 // CHECK: {{.+}}: 450 // CHECK: [[S_ARR_EL_PAST:%.+]] = phi [[S_FLOAT_TY]]* 451 // CHECK: [[S_ARR_PRIV_ITEM:%.+]] = getelementptr {{.+}}, {{.+}} [[S_ARR_EL_PAST]], 452 // CHECK: call {{.*}} [[S_FLOAT_TY_DEF_DESTR]]([[S_FLOAT_TY]]* [[S_ARR_PRIV_ITEM]]) 453 454 // CHECK: ret void 455 456 // template tmain with S_INT_TY 457 // CHECK-LABEL: define{{.*}} i{{[0-9]+}} @{{.+}}tmain{{.+}}() 458 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]], 459 // CHECK: call {{.*}} [[S_INT_TY_DEF_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]]) 460 // CHECK: call i{{[0-9]+}} @__tgt_target_teams( 461 // CHECK: call void [[OFFLOAD_FUN_0:@.+]]( 462 // CHECK: call {{.*}} [[S_INT_TY_DEF_DESTR:@.+]]([[S_INT_TY]]* [[TEST]]) 463 464 // CHECK: define{{.+}} [[OFFLOAD_FUN_0]](i{{[0-9]+}} [[T_VAR_IN:%.+]], [2 x i{{[0-9]+}}]* {{.+}} [[VEC_IN:%.+]], [2 x [[S_INT_TY]]]* {{.+}} [[S_ARR_IN:%.+]], [[S_INT_TY]]* {{.+}} [[VAR_IN:%.+]]) 465 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_teams(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, i{{[0-9]+}}*, [2 x i{{[0-9]+}}]*, [2 x [[S_INT_TY]]]*, [[S_INT_TY]]*)* [[OMP_OUTLINED_0:@.+]] to void 466 // CHECK: ret 467 468 // CHECK: define internal void [[OMP_OUTLINED_0]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i{{[0-9]+}}*{{.+}} [[T_VAR_IN:%.+]], [2 x i{{[0-9]+}}]* {{.+}} [[VEC_IN:%.+]], [2 x [[S_INT_TY]]]* {{.+}} [[S_ARR_IN:%.+]], [[S_INT_TY]]* {{.+}} [[VAR_IN:%.+]]) 469 470 // CHECK: alloca i{{[0-9]+}}*, 471 // CHECK: alloca i{{[0-9]+}}*, 472 // addr alloca's 473 // CHECK: [[T_VAR_ADDR:%.+]] = alloca i{{[0-9]+}}*, 474 // CHECK: [[VEC_ADDR:%.+]] = alloca [2 x i{{[0-9]+}}]*, 475 // CHECK: [[S_ARR_ADDR:%.+]] = alloca [2 x [[S_INT_TY]]]*, 476 // CHECK: [[VAR_ADDR:%.+]] = alloca [[S_INT_TY]]*, 477 // CHECK: [[TMP:%.+]] = alloca [[S_INT_TY]]*, 478 479 // skip loop alloca's 480 // CHECK: [[OMP_IV:.omp.iv+]] = alloca i{{[0-9]+}}, 481 // CHECK: [[OMP_LB:.omp.comb.lb+]] = alloca i{{[0-9]+}}, 482 // CHECK: [[OMP_UB:.omp.comb.ub+]] = alloca i{{[0-9]+}}, 483 // CHECK: [[OMP_ST:.omp.stride+]] = alloca i{{[0-9]+}}, 484 // CHECK: [[OMP_IS_LAST:.omp.is_last+]] = alloca i{{[0-9]+}}, 485 486 // private alloca's 487 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 488 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}], 489 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]], 490 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], 491 // CHECK: [[TMP_PRIV:%.+]] = alloca [[S_INT_TY]]*, 492 493 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_REF:%.+]] 494 495 // init addr alloca's with input values 496 // CHECK-DAG: store {{.+}} [[T_VAR_IN]], {{.+}}* [[T_VAR_ADDR]], 497 // CHECK-DAG: store {{.+}} [[VEC_IN]], {{.+}} [[VEC_ADDR]], 498 // CHECK-DAG: store {{.+}} [[S_ARR_IN]], {{.+}} [[S_ARR_ADDR]], 499 // CHECK-DAG: store {{.+}} [[VAR_IN]], {{.+}} [[VAR_ADDR]], 500 501 // init private alloca's with addr alloca's 502 // t-var 503 // CHECK-DAG: [[T_VAR_ADDR_REF:%.+]] = load {{.+}}*, {{.+}}** [[T_VAR_ADDR]], 504 // CHECK-DAG: [[T_VAR_ADDR_VAL:%.+]] = load {{.+}}, {{.+}}* [[T_VAR_ADDR_REF]], 505 // CHECK-DAG: store {{.+}} [[T_VAR_ADDR_VAL]], {{.+}} [[T_VAR_PRIV]], 506 507 // vec 508 // CHECK-DAG: [[VEC_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[VEC_ADDR]], 509 // CHECK-DAG: [[VEC_PRIV_BCAST:%.+]] = bitcast {{.+}} [[VEC_PRIV]] to 510 // CHECK-DAG: [[VEC_ADDR_BCAST:%.+]] = bitcast {{.+}} [[VEC_ADDR_VAL]] to 511 // CHECK-DAG: call void @llvm.memcpy{{.+}}({{.+}}* align {{[0-9]+}} [[VEC_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[VEC_ADDR_BCAST]], 512 513 // s_arr 514 // CHECK-DAG: [[S_ARR_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[S_ARR_ADDR]], 515 // CHECK-DAG: [[S_ARR_BGN:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_PRIV]], 516 // CHECK-DAG: [[S_ARR_ADDR_BCAST:%.+]] = bitcast {{.+}}* [[S_ARR_ADDR_VAL]] to 517 // CHECK-DAG: [[S_ARR_BGN_GEP:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_BGN]], 518 // CHECK-DAG: [[S_ARR_EMPTY:%.+]] = icmp {{.+}} [[S_ARR_BGN]], [[S_ARR_BGN_GEP]] 519 // CHECK-DAG: br {{.+}} [[S_ARR_EMPTY]], label %[[CPY_DONE:.+]], label %[[CPY_BODY:.+]] 520 // CHECK-DAG: [[CPY_BODY]]: 521 // CHECK-DAG: call void @llvm.memcpy{{.+}}( 522 // CHECK-DAG: [[CPY_DONE]]: 523 524 // var 525 // CHECK-DAG: [[TMP_REF:%.+]] = load {{.+}}*, {{.+}}* [[TMP]], 526 // CHECK-DAG: [[VAR_PRIV_BCAST:%.+]] = bitcast {{.+}}* [[VAR_PRIV]] to 527 // CHECK-DAG: [[TMP_REF_BCAST:%.+]] = bitcast {{.+}}* [[TMP_REF]] to 528 // CHECK-DAG: call void @llvm.memcpy.{{.+}}({{.+}}* align {{[0-9]+}} [[VAR_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[TMP_REF_BCAST]], 529 // CHECK-DAG: store {{.+}}* [[VAR_PRIV]], {{.+}}** [[TMP_PRIV]], 530 531 // CHECK: call void @__kmpc_for_static_init_4( 532 // pass private alloca's to fork 533 // CHECK-DAG: [[T_VAR_PRIV_VAL:%.+]] = load {{.+}}, {{.+}}* [[T_VAR_PRIV]], 534 // not dag to distinguish with S_VAR_CAST 535 // CHECK-64: [[T_VAR_CAST_CONV:%.+]] = bitcast {{.+}}* [[T_VAR_CAST:%.+]] to 536 // CHECK-64-DAG: store {{.+}} [[T_VAR_PRIV_VAL]], {{.+}} [[T_VAR_CAST_CONV]], 537 // CHECK-32: store {{.+}} [[T_VAR_PRIV_VAL]], {{.+}} [[T_VAR_CAST:%.+]], 538 // CHECK-DAG: [[T_VAR_CAST_VAL:%.+]] = load {{.+}}, {{.+}}* [[T_VAR_CAST]], 539 // CHECK-DAG: [[TMP_PRIV_VAL:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** [[TMP_PRIV]], 540 // CHECK: call{{.+}} @__kmpc_fork_call({{.+}}, {{.+}}, {{.+}}[[OMP_PARFOR_OUTLINED_0:@.+]] to void ({{.+}})*), {{.+}}, {{.+}}, [2 x i{{[0-9]+}}]* [[VEC_PRIV]], i{{[0-9]+}} [[T_VAR_CAST_VAL]], [2 x [[S_INT_TY]]]* [[S_ARR_PRIV]], [[S_INT_TY]]* [[TMP_PRIV_VAL]]) 541 // CHECK: call void @__kmpc_for_static_fini( 542 543 // call destructors: var.. 544 // CHECK-DAG: call {{.+}} [[S_INT_TY_DEF_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) 545 546 // ..and s_arr 547 // CHECK: {{.+}}: 548 // CHECK: [[S_ARR_EL_PAST:%.+]] = phi [[S_INT_TY]]* 549 // CHECK: [[S_ARR_PRIV_ITEM:%.+]] = getelementptr {{.+}}, {{.+}} [[S_ARR_EL_PAST]], 550 // CHECK: call {{.*}} [[S_INT_TY_DEF_DESTR]]([[S_INT_TY]]* [[S_ARR_PRIV_ITEM]]) 551 552 // CHECK: ret void 553 554 // By OpenMP specifications, 'firstprivate' applies to both distribute and parallel for. 555 // However, the support for 'firstprivate' of 'parallel' is only used when 'parallel' 556 // is found alone. Therefore we only have one 'firstprivate' support for 'parallel for' 557 // in combination 558 // CHECK: define internal void [[OMP_PARFOR_OUTLINED_0]]({{.+}}, {{.+}}, {{.+}}, {{.+}}, [2 x i{{[0-9]+}}]* {{.+}} [[VEC_IN:%.+]], i{{[0-9]+}} [[T_VAR_IN:%.+]], [2 x [[S_INT_TY]]]* {{.+}} [[S_ARR_IN:%.+]], [[S_INT_TY]]* {{.+}} [[VAR_IN:%.+]]) 559 560 // addr alloca's 561 // CHECK: [[VEC_ADDR:%.+]] = alloca [2 x i{{[0-9]+}}]*, 562 // CHECK: [[T_VAR_ADDR:%.+]] = alloca i{{[0-9]+}}, 563 // CHECK: [[S_ARR_ADDR:%.+]] = alloca [2 x [[S_INT_TY]]]*, 564 // CHECK: [[VAR_ADDR:%.+]] = alloca [[S_INT_TY]]*, 565 566 // skip loop alloca's 567 // CHECK: [[OMP_IV:.omp.iv+]] = alloca i{{[0-9]+}}, 568 // CHECK: [[OMP_LB:.omp.lb+]] = alloca i{{[0-9]+}}, 569 // CHECK: [[OMP_UB:.omp.ub+]] = alloca i{{[0-9]+}}, 570 // CHECK: [[OMP_ST:.omp.stride+]] = alloca i{{[0-9]+}}, 571 // CHECK: [[OMP_IS_LAST:.omp.is_last+]] = alloca i{{[0-9]+}}, 572 573 // private alloca's 574 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}], 575 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]], 576 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], 577 // CHECK: [[TMP_PRIV:%.+]] = alloca [[S_INT_TY]]*, 578 579 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_REF:%.+]] 580 581 // init addr alloca's with input values 582 // CHECK-DAG: store {{.+}} [[VEC_IN]], {{.+}} [[VEC_ADDR]], 583 // CHECK-DAG: store {{.+}} [[T_VAR_IN]], {{.+}}* [[T_VAR_ADDR]], 584 // CHECK-DAG: store {{.+}} [[S_ARR_IN]], {{.+}} [[S_ARR_ADDR]], 585 // CHECK-DAG: store {{.+}} [[VAR_IN]], {{.+}} [[VAR_ADDR]], 586 587 // init private alloca's with addr alloca's 588 // vec 589 // CHECK-DAG: [[VEC_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[VEC_ADDR]], 590 // CHECK-DAG: [[VEC_PRIV_BCAST:%.+]] = bitcast {{.+}} [[VEC_PRIV]] to 591 // CHECK-DAG: [[VEC_ADDR_BCAST:%.+]] = bitcast {{.+}} [[VEC_ADDR_VAL]] to 592 // CHECK-DAG: call void @llvm.memcpy{{.+}}({{.+}}* align {{[0-9]+}} [[VEC_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[VEC_ADDR_BCAST]], 593 594 // s_arr 595 // CHECK-DAG: [[S_ARR_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[S_ARR_ADDR]], 596 // CHECK-DAG: [[S_ARR_BGN:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_PRIV]], 597 // CHECK-DAG: [[S_ARR_ADDR_BCAST:%.+]] = bitcast {{.+}}* [[S_ARR_ADDR_VAL]] to 598 // CHECK-DAG: [[S_ARR_BGN_GEP:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_BGN]], 599 // CHECK-DAG: [[S_ARR_EMPTY:%.+]] = icmp {{.+}} [[S_ARR_BGN]], [[S_ARR_BGN_GEP]] 600 // CHECK-DAG: br {{.+}} [[S_ARR_EMPTY]], label %[[CPY_DONE:.+]], label %[[CPY_BODY:.+]] 601 // CHECK-DAG: [[CPY_BODY]]: 602 // CHECK-DAG: call void @llvm.memcpy{{.+}}( 603 // CHECK-DAG: [[CPY_DONE]]: 604 605 // var 606 // CHECK-DAG: [[VAR_ADDR_REF:%.+]] = load {{.+}}*, {{.+}}* [[VAR_ADDR]], 607 // CHECK-DAG: [[VAR_PRIV_BCAST:%.+]] = bitcast {{.+}}* [[VAR_PRIV]] to 608 // CHECK-DAG: [[VAR_ADDR_BCAST:%.+]] = bitcast {{.+}}* [[VAR_ADDR_REF]] to 609 // CHECK-DAG: call void @llvm.memcpy.{{.+}}({{.+}}* align {{[0-9]+}} [[VAR_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[VAR_ADDR_BCAST]], 610 // CHECK-DAG: store {{.+}}* [[VAR_PRIV]], {{.+}}** [[TMP_PRIV]], 611 612 // CHECK: call void @__kmpc_for_static_init_4( 613 // CHECK: call void @__kmpc_for_static_fini( 614 615 // call destructors: var.. 616 // CHECK-DAG: call {{.+}} [[S_INT_TY_DEF_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) 617 618 // ..and s_arr 619 // CHECK: {{.+}}: 620 // CHECK: [[S_ARR_EL_PAST:%.+]] = phi [[S_INT_TY]]* 621 // CHECK: [[S_ARR_PRIV_ITEM:%.+]] = getelementptr {{.+}}, {{.+}} [[S_ARR_EL_PAST]], 622 // CHECK: call {{.*}} [[S_INT_TY_DEF_DESTR]]([[S_INT_TY]]* [[S_ARR_PRIV_ITEM]]) 623 624 // CHECK: ret void 625 626 #endif 627