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 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 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 // LAMBDA-DAG: [[G_CONV:%.+]] = load {{.+}}*, {{.+}}** [[G_ADDR]] 104 // LAMBDA-DAG: [[G1_CONV:%.+]] = load {{.+}}*, {{.+}}** [[G1_ADDR]] 105 // LAMBDA-DAG: [[SVAR_CONV:%.+]] = load {{.+}}*, {{.+}}** [[SVAR_ADDR]] 106 // LAMBDA-DAG: [[SFVAR_CONV:%.+]] = load {{.+}}*, {{.+}}** [[SFVAR_ADDR]] 107 108 // init private alloca's with addr alloca's 109 // g 110 // LAMBDA-DAG: [[G_ADDR_VAL:%.+]] = load {{.+}}, {{.+}}* [[G_CONV]], 111 // LAMBDA-DAG: store {{.+}} [[G_ADDR_VAL]], {{.+}}* [[G_PRIV]], 112 113 // g1 114 // LAMBDA-DAG: [[TMP_REF:%.+]] = load {{.+}}*, {{.+}}** [[G1_REF]], 115 // LAMBDA-DAG: [[TMP_VAL:%.+]] = load {{.+}}, {{.+}}* [[TMP_REF]], 116 // LAMBDA-DAG: store {{.+}} [[TMP_VAL]], {{.+}}* [[G1_PRIV]] 117 // LAMBDA-DAG: store {{.+}}* [[G1_PRIV]], {{.+}}** [[TMP_PRIV]], 118 119 // svar 120 // LAMBDA-DAG: [[SVAR_VAL:%.+]] = load {{.+}}, {{.+}}* [[SVAR_CONV]], 121 // LAMBDA-DAG: store {{.+}} [[SVAR_VAL]], {{.+}}* [[SVAR_PRIV]], 122 123 // sfvar 124 // LAMBDA-DAG: [[SFVAR_VAL:%.+]] = load {{.+}}, {{.+}}* [[SFVAR_CONV]], 125 // LAMBDA-DAG: store {{.+}} [[SFVAR_VAL]], {{.+}}* [[SFVAR_PRIV]], 126 127 // LAMBDA: call {{.*}}void @__kmpc_for_static_init_4( 128 // pass firstprivate parameters to parallel outlined function 129 // g 130 // LAMBDA-64-DAG: [[G_PRIV_VAL:%.+]] = load {{.+}}, {{.+}}* [[G_PRIV]], 131 // LAMBDA-64: [[G_CAST_CONV:%.+]] = bitcast {{.+}}* [[G_CAST:%.+]] to 132 // LAMBDA-64-DAG: store {{.+}} [[G_PRIV_VAL]], {{.+}}* [[G_CAST_CONV]], 133 // LAMBDA-64-DAG: [[G_PAR:%.+]] = load {{.+}}, {{.+}}* [[G_CAST]], 134 135 // g1 136 // LAMBDA-DAG: [[TMP_PRIV_VAL:%.+]] = load {{.+}}, {{.+}}* [[TMP_PRIV]], 137 // LAMBDA-DAG: [[G1_PRIV_VAL:%.+]] = load {{.+}}, {{.+}}* [[TMP_PRIV_VAL]], 138 // LAMBDA: [[G1_CAST_CONV:%.+]] = bitcast {{.+}}* [[G1_CAST:%.+]] to 139 // LAMBDA-DAG: store {{.+}} [[G1_PRIV_VAL]], {{.+}}* [[G1_CAST_CONV]], 140 // LAMBDA-DAG: [[G1_PAR:%.+]] = load {{.+}}, {{.+}}* [[G1_CAST]], 141 142 // svar 143 // LAMBDA: [[SVAR_VAL:%.+]] = load {{.+}}, {{.+}}* [[SVAR_PRIV]], 144 // LAMBDA-64-DAG: [[SVAR_CAST_CONV:%.+]] = bitcast {{.+}}* [[SVAR_CAST:%.+]] to 145 // LAMBDA-64-DAG: store {{.+}} [[SVAR_VAL]], {{.+}}* [[SVAR_CAST_CONV]], 146 // LAMBDA-32-DAG: store {{.+}} [[SVAR_VAL]], {{.+}}* [[SVAR_CAST:%.+]], 147 // LAMBDA-DAG: [[SVAR_PAR:%.+]] = load {{.+}}, {{.+}}* [[SVAR_CAST]], 148 149 // sfvar 150 // LAMBDA: [[SFVAR_VAL:%.+]] = load {{.+}}, {{.+}}* [[SFVAR_PRIV]], 151 // LAMBDA-DAG: [[SFVAR_CAST_CONV:%.+]] = bitcast {{.+}}* [[SFVAR_CAST:%.+]] to 152 // LAMBDA-DAG: store {{.+}} [[SFVAR_VAL]], {{.+}}* [[SFVAR_CAST_CONV]], 153 // LAMBDA-DAG: [[SFVAR_PAR:%.+]] = load {{.+}}, {{.+}}* [[SFVAR_CAST]], 154 155 // LAMBDA-64: call{{.+}} @__kmpc_fork_call({{.+}}, {{.+}}, {{.+}}[[OMP_PARFOR_OUTLINED:@.+]] to void ({{.+}})*), {{.+}}, {{.+}}, {{.+}} [[G_PAR]], {{.+}} [[G1_PAR]], {{.+}} [[SVAR_PAR]], {{.+}} [[SFVAR_PAR]]) 156 // LAMBDA-32: call{{.+}} @__kmpc_fork_call({{.+}}, {{.+}}, {{.+}}[[OMP_PARFOR_OUTLINED:@.+]] to void ({{.+}})*), {{.+}}, {{.+}}, {{.+}} [[G_PRIV]], {{.+}} [[G1_PAR]], {{.+}} [[SVAR_PAR]], {{.+}} [[SFVAR_PAR]]) 157 // LAMBDA: call {{.*}}void @__kmpc_for_static_fini( 158 // LAMBDA: ret void 159 160 161 // 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:%.+]]) 162 // 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:%.+]]) 163 // skip initial params 164 // LAMBDA: {{.+}} = alloca{{.+}}, 165 // LAMBDA: {{.+}} = alloca{{.+}}, 166 // LAMBDA: {{.+}} = alloca{{.+}}, 167 // LAMBDA: {{.+}} = alloca{{.+}}, 168 169 // addr alloca's 170 // LAMBDA-64: [[G_ADDR:%.+]] = alloca i{{[0-9]+}}, 171 // LAMBDA-32: [[G_ADDR:%.+]] = alloca double*, 172 // LAMBDA: [[G1_ADDR:%.+]] = alloca i{{[0-9]+}}, 173 // LAMBDA: [[SVAR_ADDR:%.+]] = alloca i{{[0-9]+}}, 174 // LAMBDA: [[SFVAR_ADDR:%.+]] = alloca i{{[0-9]+}}, 175 // LAMBDA: [[G1_REF:%.+]] = alloca double*, 176 177 // private alloca's (only for 32-bit) 178 // LAMBDA-32: [[G_PRIV:%.+]] = alloca double, 179 180 // transfer input parameters into addr alloca's 181 // LAMBDA-DAG: store {{.+}} [[G_IN]], {{.+}} [[G_ADDR]], 182 // LAMBDA-DAG: store {{.+}} [[G1_IN]], {{.+}} [[G1_ADDR]], 183 // LAMBDA-DAG: store {{.+}} [[SVAR_IN]], {{.+}} [[SVAR_ADDR]], 184 // LAMBDA-DAG: store {{.+}} [[SFVAR_IN]], {{.+}} [[SFVAR_ADDR]], 185 186 // prepare parameters for lambda 187 // g 188 // LAMBDA-64-DAG: [[G_CONV:%.+]] = bitcast {{.+}}* [[G_ADDR]] to 189 // LAMBDA-32-DAG: [[G_ADDR_REF:%.+]] = load {{.+}}*, {{.+}}** [[G_ADDR]] 190 // LAMBDA-32-DAG: [[G_ADDR_VAL:%.+]] = load {{.+}}, {{.+}}* [[G_ADDR_REF]], 191 // LAMBDA-32-DAG: store {{.+}} [[G_ADDR_VAL]], {{.+}}* [[G_PRIV]], 192 193 // g1 194 // LAMBDA-DAG: [[G1_CONV:%.+]] = bitcast {{.+}}* [[G1_ADDR]] to 195 // LAMBDA-DAG: store {{.+}}* [[G1_CONV]], {{.+}}* [[G1_REF]], 196 197 // svar 198 // LAMBDA-64-DAG: [[SVAR_CONV:%.+]] = bitcast {{.+}}* [[SVAR_ADDR]] to 199 200 // sfvar 201 // LAMBDA-DAG: [[SFVAR_CONV:%.+]] = bitcast {{.+}}* [[SFVAR_ADDR]] to 202 203 // LAMBDA: call {{.*}}void @__kmpc_for_static_init_4( 204 g = 1; 205 g1 = 1; 206 svar = 3; 207 sfvar = 4.0; 208 // LAMBDA-64: store double 1.0{{.+}}, double* [[G_CONV]], 209 // LAMBDA-32: store double 1.0{{.+}}, double* [[G_PRIV]], 210 // LAMBDA: [[G1_REF_REF:%.+]] = load {{.+}}*, {{.+}}** [[G1_REF]], 211 // LAMBDA: store {{.+}} 1.0{{.+}}, {{.+}}* [[G1_REF_REF]], 212 // LAMBDA-64: store {{.+}} 3, {{.+}}* [[SVAR_CONV]], 213 // LAMBDA-32: store {{.+}} 3, {{.+}}* [[SVAR_ADDR]], 214 // LAMBDA: store {{.+}} 4.0{{.+}}, {{.+}}* [[SFVAR_CONV]], 215 216 // pass params to inner lambda 217 // LAMBDA: [[G_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 218 // LAMBDA-64: store double* [[G_CONV]], double** [[G_PRIVATE_ADDR_REF]], 219 // LAMBDA-32: store double* [[G_PRIV]], double** [[G_PRIVATE_ADDR_REF]], 220 // LAMBDA: [[G1_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 1 221 // LAMBDA: [[G1_REF_REF:%.+]] = load double*, double** [[G1_REF]], 222 // LAMBDA: store double* [[G1_REF_REF]], double** [[G1_PRIVATE_ADDR_REF]], 223 // LAMBDA: [[SVAR_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 2 224 // LAMBDA-64: store i{{[0-9]+}}* [[SVAR_CONV]], i{{[0-9]+}}** [[SVAR_PRIVATE_ADDR_REF]] 225 // LAMBDA-32: store i{{[0-9]+}}* [[SVAR_ADDR]], i{{[0-9]+}}** [[SVAR_PRIVATE_ADDR_REF]] 226 // LAMBDA: [[SFVAR_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 3 227 // LAMBDA: store float* [[SFVAR_CONV]], float** [[SFVAR_PRIVATE_ADDR_REF]] 228 // LAMBDA: call{{.*}} void [[INNER_LAMBDA:@.+]](%{{.+}}* [[ARG]]) 229 // LAMBDA: call {{.*}}void @__kmpc_for_static_fini( 230 // LAMBDA: ret void 231 [&]() { 232 // LAMBDA: define {{.+}} void [[INNER_LAMBDA]](%{{.+}}* [[ARG_PTR:%.+]]) 233 // LAMBDA: store %{{.+}}* [[ARG_PTR]], %{{.+}}** [[ARG_PTR_REF:%.+]], 234 g = 2; 235 g1 = 2; 236 svar = 4; 237 sfvar = 8.0; 238 // LAMBDA: [[ARG_PTR:%.+]] = load %{{.+}}*, %{{.+}}** [[ARG_PTR_REF]] 239 // LAMBDA: [[G_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 240 // LAMBDA: [[G_REF:%.+]] = load double*, double** [[G_PTR_REF]] 241 // LAMBDA: store double 2.0{{.+}}, double* [[G_REF]] 242 243 // LAMBDA: [[TMP_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 1 244 // LAMBDA: [[G1_REF:%.+]] = load double*, double** [[TMP_PTR_REF]] 245 // LAMBDA: store double 2.0{{.+}}, double* [[G1_REF]], 246 // LAMBDA: [[SVAR_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 2 247 // LAMBDA: [[SVAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[SVAR_PTR_REF]] 248 // LAMBDA: store i{{[0-9]+}} 4, i{{[0-9]+}}* [[SVAR_REF]] 249 // LAMBDA: [[SFVAR_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 3 250 // LAMBDA: [[SFVAR_REF:%.+]] = load float*, float** [[SFVAR_PTR_REF]] 251 // LAMBDA: store float 8.0{{.+}}, float* [[SFVAR_REF]] 252 }(); 253 } 254 }(); 255 return 0; 256 #else 257 S<float> test; 258 int t_var = 0; 259 int vec[] = {1, 2}; 260 S<float> s_arr[] = {1, 2}; 261 S<float> &var = test; 262 263 #pragma omp target 264 #pragma omp teams 265 #pragma omp distribute parallel for firstprivate(t_var, vec, s_arr, s_arr, var, var, svar) 266 for (int i = 0; i < 2; ++i) { 267 vec[i] = t_var; 268 s_arr[i] = var; 269 } 270 return tmain<int>(); 271 #endif 272 } 273 274 // CHECK-LABEL: define{{.*}} i{{[0-9]+}} @main() 275 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]], 276 // CHECK: call {{.*}} [[S_FLOAT_TY_DEF_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]]) 277 // CHECK: call i{{[0-9]+}} @__tgt_target_teams( 278 // CHECK: call void [[OFFLOAD_FUN_0:@.+]]( 279 // CHECK: call {{.*}} [[S_FLOAT_TY_DEF_DESTR:@.+]]([[S_FLOAT_TY]]* [[TEST]]) 280 281 // 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:%.+]]) 282 // 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 283 // CHECK: ret 284 285 // 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:%.+]]) 286 287 // CHECK: alloca i{{[0-9]+}}*, 288 // CHECK: alloca i{{[0-9]+}}*, 289 // addr alloca's 290 // CHECK: [[T_VAR_ADDR:%.+]] = alloca i{{[0-9]+}}*, 291 // CHECK: [[VEC_ADDR:%.+]] = alloca [2 x i{{[0-9]+}}]*, 292 // CHECK: [[S_ARR_ADDR:%.+]] = alloca [2 x [[S_FLOAT_TY]]]*, 293 // CHECK: [[VAR_ADDR:%.+]] = alloca [[S_FLOAT_TY]]*, 294 // CHECK: [[SVAR_ADDR:%.+]] = alloca i{{[0-9]+}}*, 295 // CHECK: [[TMP:%.+]] = alloca [[S_FLOAT_TY]]*, 296 297 // skip loop alloca's 298 // CHECK: [[OMP_IV:.omp.iv+]] = alloca i{{[0-9]+}}, 299 // CHECK: [[OMP_LB:.omp.comb.lb+]] = alloca i{{[0-9]+}}, 300 // CHECK: [[OMP_UB:.omp.comb.ub+]] = alloca i{{[0-9]+}}, 301 // CHECK: [[OMP_ST:.omp.stride+]] = alloca i{{[0-9]+}}, 302 // CHECK: [[OMP_IS_LAST:.omp.is_last+]] = alloca i{{[0-9]+}}, 303 304 // private alloca's 305 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 306 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}], 307 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]], 308 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 309 // CHECK: [[TMP_PRIV:%.+]] = alloca [[S_FLOAT_TY]]*, 310 // CHECK: [[SVAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 311 312 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_REF:%.+]] 313 314 // init addr alloca's with input values 315 // CHECK-DAG: store {{.+}} [[T_VAR_IN]], {{.+}}* [[T_VAR_ADDR]], 316 // CHECK-DAG: store {{.+}} [[VEC_IN]], {{.+}} [[VEC_ADDR]], 317 // CHECK-DAG: store {{.+}} [[S_ARR_IN]], {{.+}} [[S_ARR_ADDR]], 318 // CHECK-DAG: store {{.+}} [[VAR_IN]], {{.+}} [[VAR_ADDR]], 319 // CHECK-DAG: store {{.+}} [[SVAR_IN]], {{.+}} [[SVAR_ADDR]], 320 321 // init private alloca's with addr alloca's 322 // t-var 323 // CHECK-DAG: [[T_VAR_REF:%.+]] = load {{.+}}, {{.+}}** [[T_VAR_ADDR]], 324 // CHECK-DAG: [[T_VAR_ADDR_VAL:%.+]] = load {{.+}}, {{.+}}* [[T_VAR_REF]], 325 // CHECK-DAG: store {{.+}} [[T_VAR_ADDR_VAL]], {{.+}} [[T_VAR_PRIV]], 326 327 // vec 328 // CHECK-DAG: [[VEC_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[VEC_ADDR]], 329 // CHECK-DAG: [[VEC_PRIV_BCAST:%.+]] = bitcast {{.+}} [[VEC_PRIV]] to 330 // CHECK-DAG: [[VEC_ADDR_BCAST:%.+]] = bitcast {{.+}} [[VEC_ADDR_VAL]] to 331 // CHECK-DAG: call void @llvm.memcpy{{.+}}({{.+}}* align {{[0-9]+}} [[VEC_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[VEC_ADDR_BCAST]], 332 333 // s_arr 334 // CHECK-DAG: [[S_ARR_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[S_ARR_ADDR]], 335 // CHECK-DAG: [[S_ARR_BGN:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_PRIV]], 336 // CHECK-DAG: [[S_ARR_ADDR_BCAST:%.+]] = bitcast {{.+}}* [[S_ARR_ADDR_VAL]] to 337 // CHECK-DAG: [[S_ARR_BGN_GEP:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_BGN]], 338 // CHECK-DAG: [[S_ARR_EMPTY:%.+]] = icmp {{.+}} [[S_ARR_BGN]], [[S_ARR_BGN_GEP]] 339 // CHECK-DAG: br {{.+}} [[S_ARR_EMPTY]], label %[[CPY_DONE:.+]], label %[[CPY_BODY:.+]] 340 // CHECK-DAG: [[CPY_BODY]]: 341 // CHECK-DAG: call void @llvm.memcpy{{.+}}( 342 // CHECK-DAG: [[CPY_DONE]]: 343 344 // var 345 // CHECK-DAG: [[TMP_REF:%.+]] = load {{.+}}*, {{.+}}* [[TMP]], 346 // CHECK-DAG: [[VAR_PRIV_BCAST:%.+]] = bitcast {{.+}}* [[VAR_PRIV]] to 347 // CHECK-DAG: [[TMP_REF_BCAST:%.+]] = bitcast {{.+}}* [[TMP_REF]] to 348 // CHECK-DAG: call void @llvm.memcpy.{{.+}}({{.+}}* align {{[0-9]+}} [[VAR_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[TMP_REF_BCAST]], 349 // CHECK-DAG: store {{.+}}* [[VAR_PRIV]], {{.+}}** [[TMP_PRIV]], 350 351 // svar 352 // CHECK-DAG: [[SVAR_REF:%.+]] = load {{.+}}*, {{.+}}** [[SVAR_ADDR]], 353 // CHECK-DAG: [[SVAR:%.+]] = load {{.+}}, {{.+}}* [[SVAR_REF]], 354 // CHECK-DAG: store {{.+}} [[SVAR]], {{.+}}* [[SVAR_PRIV]], 355 356 // CHECK: call void @__kmpc_for_static_init_4( 357 // pass private alloca's to fork 358 // CHECK-DAG: [[T_VAR_PRIV_VAL:%.+]] = load {{.+}}, {{.+}}* [[T_VAR_PRIV]], 359 // not dag to distinguish with S_VAR_CAST 360 // CHECK-64: [[T_VAR_CAST_CONV:%.+]] = bitcast {{.+}}* [[T_VAR_CAST:%.+]] to 361 // CHECK-64-DAG: store {{.+}} [[T_VAR_PRIV_VAL]], {{.+}} [[T_VAR_CAST_CONV]], 362 // CHECK-32: store {{.+}} [[T_VAR_PRIV_VAL]], {{.+}} [[T_VAR_CAST:%.+]], 363 // CHECK-DAG: [[T_VAR_CAST_VAL:%.+]] = load {{.+}}, {{.+}}* [[T_VAR_CAST]], 364 // CHECK-DAG: [[TMP_PRIV_VAL:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[TMP_PRIV]], 365 // CHECK-DAG: [[SVAR_PRIV_VAL:%.+]] = load {{.+}}, {{.+}}* [[SVAR_PRIV]], 366 // CHECK-64-DAG: [[SVAR_CAST_CONV:%.+]] = bitcast {{.+}}* [[SVAR_CAST:%.+]] to 367 // CHECK-64-DAG: store {{.+}} [[SVAR_PRIV_VAL]], {{.+}}* [[SVAR_CAST_CONV]], 368 // CHECK-32-DAG: store {{.+}} [[SVAR_PRIV_VAL]], {{.+}}* [[SVAR_CAST:%.+]], 369 // CHECK-DAG: [[SVAR_CAST_VAL:%.+]] = load {{.+}}, {{.+}}* [[SVAR_CAST]], 370 // 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]]) 371 // CHECK: call void @__kmpc_for_static_fini( 372 373 // call destructors: var.. 374 // CHECK-DAG: call {{.+}} [[S_FLOAT_TY_DEF_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 375 376 // ..and s_arr 377 // CHECK: {{.+}}: 378 // CHECK: [[S_ARR_EL_PAST:%.+]] = phi [[S_FLOAT_TY]]* 379 // CHECK: [[S_ARR_PRIV_ITEM:%.+]] = getelementptr {{.+}}, {{.+}} [[S_ARR_EL_PAST]], 380 // CHECK: call {{.*}} [[S_FLOAT_TY_DEF_DESTR]]([[S_FLOAT_TY]]* [[S_ARR_PRIV_ITEM]]) 381 382 // CHECK: ret void 383 384 // By OpenMP specifications, 'firstprivate' applies to both distribute and parallel for. 385 // However, the support for 'firstprivate' of 'parallel' is only used when 'parallel' 386 // is found alone. Therefore we only have one 'firstprivate' support for 'parallel for' 387 // in combination 388 // 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:%.+]]) 389 390 // addr alloca's 391 // CHECK: [[VEC_ADDR:%.+]] = alloca [2 x i{{[0-9]+}}]*, 392 // CHECK: [[T_VAR_ADDR:%.+]] = alloca i{{[0-9]+}}, 393 // CHECK: [[S_ARR_ADDR:%.+]] = alloca [2 x [[S_FLOAT_TY]]]*, 394 // CHECK: [[VAR_ADDR:%.+]] = alloca [[S_FLOAT_TY]]*, 395 // CHECK: [[SVAR_ADDR:%.+]] = alloca i{{[0-9]+}}, 396 397 // skip loop alloca's 398 // CHECK: [[OMP_IV:.omp.iv+]] = alloca i{{[0-9]+}}, 399 // CHECK: [[OMP_LB:.omp.lb+]] = alloca i{{[0-9]+}}, 400 // CHECK: [[OMP_UB:.omp.ub+]] = alloca i{{[0-9]+}}, 401 // CHECK: [[OMP_ST:.omp.stride+]] = alloca i{{[0-9]+}}, 402 // CHECK: [[OMP_IS_LAST:.omp.is_last+]] = alloca i{{[0-9]+}}, 403 404 // private alloca's 405 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}], 406 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]], 407 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 408 // CHECK: [[TMP_PRIV:%.+]] = alloca [[S_FLOAT_TY]]*, 409 410 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_REF:%.+]] 411 412 // init addr alloca's with input values 413 // CHECK-DAG: store {{.+}} [[VEC_IN]], {{.+}} [[VEC_ADDR]], 414 // CHECK-DAG: store {{.+}} [[T_VAR_IN]], {{.+}}* [[T_VAR_ADDR]], 415 // CHECK-DAG: store {{.+}} [[S_ARR_IN]], {{.+}} [[S_ARR_ADDR]], 416 // CHECK-DAG: store {{.+}} [[VAR_IN]], {{.+}} [[VAR_ADDR]], 417 // CHECK-DAG: store {{.+}} [[SVAR_IN]], {{.+}} [[SVAR_ADDR]], 418 419 // init private alloca's with addr alloca's 420 // vec 421 // CHECK-DAG: [[VEC_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[VEC_ADDR]], 422 // CHECK-DAG: [[VEC_PRIV_BCAST:%.+]] = bitcast {{.+}} [[VEC_PRIV]] to 423 // CHECK-DAG: [[VEC_ADDR_BCAST:%.+]] = bitcast {{.+}} [[VEC_ADDR_VAL]] to 424 // CHECK-DAG: call void @llvm.memcpy{{.+}}({{.+}}* align {{[0-9]+}} [[VEC_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[VEC_ADDR_BCAST]], 425 426 // s_arr 427 // CHECK-DAG: [[S_ARR_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[S_ARR_ADDR]], 428 // CHECK-DAG: [[S_ARR_BGN:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_PRIV]], 429 // CHECK-DAG: [[S_ARR_ADDR_BCAST:%.+]] = bitcast {{.+}}* [[S_ARR_ADDR_VAL]] to 430 // CHECK-DAG: [[S_ARR_BGN_GEP:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_BGN]], 431 // CHECK-DAG: [[S_ARR_EMPTY:%.+]] = icmp {{.+}} [[S_ARR_BGN]], [[S_ARR_BGN_GEP]] 432 // CHECK-DAG: br {{.+}} [[S_ARR_EMPTY]], label %[[CPY_DONE:.+]], label %[[CPY_BODY:.+]] 433 // CHECK-DAG: [[CPY_BODY]]: 434 // CHECK-DAG: call void @llvm.memcpy{{.+}}( 435 // CHECK-DAG: [[CPY_DONE]]: 436 437 // var 438 // CHECK-DAG: [[VAR_ADDR_REF:%.+]] = load {{.+}}*, {{.+}}* [[VAR_ADDR]], 439 // CHECK-DAG: [[VAR_PRIV_BCAST:%.+]] = bitcast {{.+}}* [[VAR_PRIV]] to 440 // CHECK-DAG: [[VAR_ADDR_BCAST:%.+]] = bitcast {{.+}}* [[VAR_ADDR_REF]] to 441 // CHECK-DAG: call void @llvm.memcpy.{{.+}}({{.+}}* align {{[0-9]+}} [[VAR_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[VAR_ADDR_BCAST]], 442 // CHECK-DAG: store {{.+}}* [[VAR_PRIV]], {{.+}}** [[TMP_PRIV]], 443 444 // CHECK: call void @__kmpc_for_static_init_4( 445 // CHECK: call void @__kmpc_for_static_fini( 446 447 // call destructors: var.. 448 // CHECK-DAG: call {{.+}} [[S_FLOAT_TY_DEF_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 449 450 // ..and s_arr 451 // CHECK: {{.+}}: 452 // CHECK: [[S_ARR_EL_PAST:%.+]] = phi [[S_FLOAT_TY]]* 453 // CHECK: [[S_ARR_PRIV_ITEM:%.+]] = getelementptr {{.+}}, {{.+}} [[S_ARR_EL_PAST]], 454 // CHECK: call {{.*}} [[S_FLOAT_TY_DEF_DESTR]]([[S_FLOAT_TY]]* [[S_ARR_PRIV_ITEM]]) 455 456 // CHECK: ret void 457 458 // template tmain with S_INT_TY 459 // CHECK-LABEL: define{{.*}} i{{[0-9]+}} @{{.+}}tmain{{.+}}() 460 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]], 461 // CHECK: call {{.*}} [[S_INT_TY_DEF_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]]) 462 // CHECK: call i{{[0-9]+}} @__tgt_target_teams( 463 // CHECK: call void [[OFFLOAD_FUN_0:@.+]]( 464 // CHECK: call {{.*}} [[S_INT_TY_DEF_DESTR:@.+]]([[S_INT_TY]]* [[TEST]]) 465 466 // 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:%.+]]) 467 // 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 468 // CHECK: ret 469 470 // 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:%.+]]) 471 472 // addr alloca's 473 // CHECK: alloca i{{[0-9]+}}*, 474 // CHECK: alloca i{{[0-9]+}}*, 475 // CHECK: [[T_VAR_ADDR:%.+]] = alloca i{{[0-9]+}}*, 476 // CHECK: [[VEC_ADDR:%.+]] = alloca [2 x i{{[0-9]+}}]*, 477 // CHECK: [[S_ARR_ADDR:%.+]] = alloca [2 x [[S_INT_TY]]]*, 478 // CHECK: [[VAR_ADDR:%.+]] = alloca [[S_INT_TY]]*, 479 // CHECK: [[TMP:%.+]] = alloca [[S_INT_TY]]*, 480 481 // skip loop alloca's 482 // CHECK: [[OMP_IV:.omp.iv+]] = alloca i{{[0-9]+}}, 483 // CHECK: [[OMP_LB:.omp.comb.lb+]] = alloca i{{[0-9]+}}, 484 // CHECK: [[OMP_UB:.omp.comb.ub+]] = alloca i{{[0-9]+}}, 485 // CHECK: [[OMP_ST:.omp.stride+]] = alloca i{{[0-9]+}}, 486 // CHECK: [[OMP_IS_LAST:.omp.is_last+]] = alloca i{{[0-9]+}}, 487 488 // private alloca's 489 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 490 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}], 491 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]], 492 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], 493 // CHECK: [[TMP_PRIV:%.+]] = alloca [[S_INT_TY]]*, 494 495 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_REF:%.+]] 496 497 // init addr alloca's with input values 498 // CHECK-DAG: store {{.+}} [[T_VAR_IN]], {{.+}}* [[T_VAR_ADDR]], 499 // CHECK-DAG: store {{.+}} [[VEC_IN]], {{.+}} [[VEC_ADDR]], 500 // CHECK-DAG: store {{.+}} [[S_ARR_IN]], {{.+}} [[S_ARR_ADDR]], 501 // CHECK-DAG: store {{.+}} [[VAR_IN]], {{.+}} [[VAR_ADDR]], 502 503 // init private alloca's with addr alloca's 504 // t-var 505 // CHECK-DAG: [[T_VAR_ADDR_REF:%.+]] = load {{.+}}*, {{.+}}** [[T_VAR_ADDR]], 506 // CHECK-DAG: [[T_VAR_ADDR_VAL:%.+]] = load {{.+}}, {{.+}}* [[T_VAR_ADDR_REF]], 507 // CHECK-DAG: store {{.+}} [[T_VAR_ADDR_VAL]], {{.+}} [[T_VAR_PRIV]], 508 509 // vec 510 // CHECK-DAG: [[VEC_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[VEC_ADDR]], 511 // CHECK-DAG: [[VEC_PRIV_BCAST:%.+]] = bitcast {{.+}} [[VEC_PRIV]] to 512 // CHECK-DAG: [[VEC_ADDR_BCAST:%.+]] = bitcast {{.+}} [[VEC_ADDR_VAL]] to 513 // CHECK-DAG: call void @llvm.memcpy{{.+}}({{.+}}* align {{[0-9]+}} [[VEC_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[VEC_ADDR_BCAST]], 514 515 // s_arr 516 // CHECK-DAG: [[S_ARR_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[S_ARR_ADDR]], 517 // CHECK-DAG: [[S_ARR_BGN:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_PRIV]], 518 // CHECK-DAG: [[S_ARR_ADDR_BCAST:%.+]] = bitcast {{.+}}* [[S_ARR_ADDR_VAL]] to 519 // CHECK-DAG: [[S_ARR_BGN_GEP:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_BGN]], 520 // CHECK-DAG: [[S_ARR_EMPTY:%.+]] = icmp {{.+}} [[S_ARR_BGN]], [[S_ARR_BGN_GEP]] 521 // CHECK-DAG: br {{.+}} [[S_ARR_EMPTY]], label %[[CPY_DONE:.+]], label %[[CPY_BODY:.+]] 522 // CHECK-DAG: [[CPY_BODY]]: 523 // CHECK-DAG: call void @llvm.memcpy{{.+}}( 524 // CHECK-DAG: [[CPY_DONE]]: 525 526 // var 527 // CHECK-DAG: [[TMP_REF:%.+]] = load {{.+}}*, {{.+}}* [[TMP]], 528 // CHECK-DAG: [[VAR_PRIV_BCAST:%.+]] = bitcast {{.+}}* [[VAR_PRIV]] to 529 // CHECK-DAG: [[TMP_REF_BCAST:%.+]] = bitcast {{.+}}* [[TMP_REF]] to 530 // CHECK-DAG: call void @llvm.memcpy.{{.+}}({{.+}}* align {{[0-9]+}} [[VAR_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[TMP_REF_BCAST]], 531 // CHECK-DAG: store {{.+}}* [[VAR_PRIV]], {{.+}}** [[TMP_PRIV]], 532 533 // CHECK: call void @__kmpc_for_static_init_4( 534 // pass private alloca's to fork 535 // CHECK-DAG: [[T_VAR_PRIV_VAL:%.+]] = load {{.+}}, {{.+}}* [[T_VAR_PRIV]], 536 // not dag to distinguish with S_VAR_CAST 537 // CHECK-64: [[T_VAR_CAST_CONV:%.+]] = bitcast {{.+}}* [[T_VAR_CAST:%.+]] to 538 // CHECK-64-DAG: store {{.+}} [[T_VAR_PRIV_VAL]], {{.+}} [[T_VAR_CAST_CONV]], 539 // CHECK-32: store {{.+}} [[T_VAR_PRIV_VAL]], {{.+}} [[T_VAR_CAST:%.+]], 540 // CHECK-DAG: [[T_VAR_CAST_VAL:%.+]] = load {{.+}}, {{.+}}* [[T_VAR_CAST]], 541 // CHECK-DAG: [[TMP_PRIV_VAL:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** [[TMP_PRIV]], 542 // 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]]) 543 // CHECK: call void @__kmpc_for_static_fini( 544 545 // call destructors: var.. 546 // CHECK-DAG: call {{.+}} [[S_INT_TY_DEF_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) 547 548 // ..and s_arr 549 // CHECK: {{.+}}: 550 // CHECK: [[S_ARR_EL_PAST:%.+]] = phi [[S_INT_TY]]* 551 // CHECK: [[S_ARR_PRIV_ITEM:%.+]] = getelementptr {{.+}}, {{.+}} [[S_ARR_EL_PAST]], 552 // CHECK: call {{.*}} [[S_INT_TY_DEF_DESTR]]([[S_INT_TY]]* [[S_ARR_PRIV_ITEM]]) 553 554 // CHECK: ret void 555 556 // By OpenMP specifications, 'firstprivate' applies to both distribute and parallel for. 557 // However, the support for 'firstprivate' of 'parallel' is only used when 'parallel' 558 // is found alone. Therefore we only have one 'firstprivate' support for 'parallel for' 559 // in combination 560 // 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:%.+]]) 561 562 // addr alloca's 563 // CHECK: [[VEC_ADDR:%.+]] = alloca [2 x i{{[0-9]+}}]*, 564 // CHECK: [[T_VAR_ADDR:%.+]] = alloca i{{[0-9]+}}, 565 // CHECK: [[S_ARR_ADDR:%.+]] = alloca [2 x [[S_INT_TY]]]*, 566 // CHECK: [[VAR_ADDR:%.+]] = alloca [[S_INT_TY]]*, 567 568 // skip loop alloca's 569 // CHECK: [[OMP_IV:.omp.iv+]] = alloca i{{[0-9]+}}, 570 // CHECK: [[OMP_LB:.omp.lb+]] = alloca i{{[0-9]+}}, 571 // CHECK: [[OMP_UB:.omp.ub+]] = alloca i{{[0-9]+}}, 572 // CHECK: [[OMP_ST:.omp.stride+]] = alloca i{{[0-9]+}}, 573 // CHECK: [[OMP_IS_LAST:.omp.is_last+]] = alloca i{{[0-9]+}}, 574 575 // private alloca's 576 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}], 577 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]], 578 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], 579 // CHECK: [[TMP_PRIV:%.+]] = alloca [[S_INT_TY]]*, 580 581 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_REF:%.+]] 582 583 // init addr alloca's with input values 584 // CHECK-DAG: store {{.+}} [[VEC_IN]], {{.+}} [[VEC_ADDR]], 585 // CHECK-DAG: store {{.+}} [[T_VAR_IN]], {{.+}}* [[T_VAR_ADDR]], 586 // CHECK-DAG: store {{.+}} [[S_ARR_IN]], {{.+}} [[S_ARR_ADDR]], 587 // CHECK-DAG: store {{.+}} [[VAR_IN]], {{.+}} [[VAR_ADDR]], 588 589 // init private alloca's with addr alloca's 590 // vec 591 // CHECK-DAG: [[VEC_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[VEC_ADDR]], 592 // CHECK-DAG: [[VEC_PRIV_BCAST:%.+]] = bitcast {{.+}} [[VEC_PRIV]] to 593 // CHECK-DAG: [[VEC_ADDR_BCAST:%.+]] = bitcast {{.+}} [[VEC_ADDR_VAL]] to 594 // CHECK-DAG: call void @llvm.memcpy{{.+}}({{.+}}* align {{[0-9]+}} [[VEC_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[VEC_ADDR_BCAST]], 595 596 // s_arr 597 // CHECK-DAG: [[S_ARR_ADDR_VAL:%.+]] = load {{.+}}*, {{.+}}** [[S_ARR_ADDR]], 598 // CHECK-DAG: [[S_ARR_BGN:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_PRIV]], 599 // CHECK-DAG: [[S_ARR_ADDR_BCAST:%.+]] = bitcast {{.+}}* [[S_ARR_ADDR_VAL]] to 600 // CHECK-DAG: [[S_ARR_BGN_GEP:%.+]] = getelementptr {{.+}}, {{.+}}* [[S_ARR_BGN]], 601 // CHECK-DAG: [[S_ARR_EMPTY:%.+]] = icmp {{.+}} [[S_ARR_BGN]], [[S_ARR_BGN_GEP]] 602 // CHECK-DAG: br {{.+}} [[S_ARR_EMPTY]], label %[[CPY_DONE:.+]], label %[[CPY_BODY:.+]] 603 // CHECK-DAG: [[CPY_BODY]]: 604 // CHECK-DAG: call void @llvm.memcpy{{.+}}( 605 // CHECK-DAG: [[CPY_DONE]]: 606 607 // var 608 // CHECK-DAG: [[VAR_ADDR_REF:%.+]] = load {{.+}}*, {{.+}}* [[VAR_ADDR]], 609 // CHECK-DAG: [[VAR_PRIV_BCAST:%.+]] = bitcast {{.+}}* [[VAR_PRIV]] to 610 // CHECK-DAG: [[VAR_ADDR_BCAST:%.+]] = bitcast {{.+}}* [[VAR_ADDR_REF]] to 611 // CHECK-DAG: call void @llvm.memcpy.{{.+}}({{.+}}* align {{[0-9]+}} [[VAR_PRIV_BCAST]], {{.+}}* align {{[0-9]+}} [[VAR_ADDR_BCAST]], 612 // CHECK-DAG: store {{.+}}* [[VAR_PRIV]], {{.+}}** [[TMP_PRIV]], 613 614 // CHECK: call void @__kmpc_for_static_init_4( 615 // CHECK: call void @__kmpc_for_static_fini( 616 617 // call destructors: var.. 618 // CHECK-DAG: call {{.+}} [[S_INT_TY_DEF_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) 619 620 // ..and s_arr 621 // CHECK: {{.+}}: 622 // CHECK: [[S_ARR_EL_PAST:%.+]] = phi [[S_INT_TY]]* 623 // CHECK: [[S_ARR_PRIV_ITEM:%.+]] = getelementptr {{.+}}, {{.+}} [[S_ARR_EL_PAST]], 624 // CHECK: call {{.*}} [[S_INT_TY_DEF_DESTR]]([[S_INT_TY]]* [[S_ARR_PRIV_ITEM]]) 625 626 // CHECK: ret void 627 628 #endif 629