1 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - -fsanitize-address-use-after-scope | FileCheck %s --check-prefix=CHECK --check-prefix=LIFETIME 2 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s 3 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s 4 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -fopenmp -fexceptions -fcxx-exceptions -debug-info-kind=line-tables-only -x c++ -emit-llvm %s -o - | FileCheck %s --check-prefix=TERM_DEBUG 5 // RUN: %clang_cc1 -main-file-name for_codegen.cpp %s -o - -emit-llvm -fprofile-instrument=clang -fprofile-instrument-path=for_codegen-test.profraw | FileCheck %s --check-prefix=PROF-INSTR-PATH 6 7 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck --check-prefix SIMD-ONLY0 %s 8 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s 9 // RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY0 %s 10 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -fopenmp-simd -fexceptions -fcxx-exceptions -debug-info-kind=line-tables-only -x c++ -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s 11 // RUN: %clang_cc1 -main-file-name for_codegen.cpp %s -o - -emit-llvm -fprofile-instrument=clang -fprofile-instrument-path=for_codegen-test.profraw | FileCheck --check-prefix SIMD-ONLY0 %s 12 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}} 13 // 14 // expected-no-diagnostics 15 #ifndef HEADER 16 #define HEADER 17 // PROF-INSTR-PATH: constant [25 x i8] c"for_codegen-test.profraw\00" 18 19 // CHECK: [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, i8* } 20 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr global %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8* 21 // CHECK-DAG: [[LOOP_LOC:@.+]] = private unnamed_addr global %{{.+}} { i32 0, i32 514, i32 0, i32 0, i8* 22 // CHECK-DAG: [[I:@.+]] = global i8 1, 23 // CHECK-DAG: [[J:@.+]] = global i8 2, 24 // CHECK-DAG: [[K:@.+]] = global i8 3, 25 26 // CHECK-LABEL: loop_with_counter_collapse 27 void loop_with_counter_collapse() { 28 // LIFETIME: call void @llvm.lifetime.end 29 // LIFETIME: call void @llvm.lifetime.end 30 // CHECK: call void @__kmpc_for_static_init_8(%struct.ident_t* @ 31 // CHECK: call void @__kmpc_for_static_fini(%struct.ident_t* @ 32 // LIFETIME: call void @llvm.lifetime.end 33 // LIFETIME: call void @llvm.lifetime.end 34 // LIFETIME: call void @llvm.lifetime.end 35 // LIFETIME: call void @llvm.lifetime.end 36 // LIFETIME: call void @llvm.lifetime.end 37 // LIFETIME: call void @llvm.lifetime.end 38 // LIFETIME: call void @llvm.lifetime.end 39 // LIFETIME: call void @llvm.lifetime.end 40 // LIFETIME: call void @llvm.lifetime.end 41 #pragma omp for collapse(2) 42 for (int i = 0; i < 4; i++) { 43 for (int j = i; j < 4; j++) { 44 } 45 } 46 } 47 // CHECK-LABEL: define {{.*void}} @{{.*}}without_schedule_clause{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 48 void without_schedule_clause(float *a, float *b, float *c, float *d) { 49 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 50 #pragma omp for nowait 51 // CHECK: call void @__kmpc_for_static_init_4([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]], i32 34, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 1) 52 // UB = min(UB, GlobalUB) 53 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 54 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423 55 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 56 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 57 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 58 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 59 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 60 // Loop header 61 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 62 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 63 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 64 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 65 for (int i = 33; i < 32000000; i += 7) { 66 // CHECK: [[LOOP1_BODY]] 67 // Start of body: calculate i from IV: 68 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 69 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7 70 // CHECK-NEXT: [[CALC_I_2:%.+]] = add nsw i32 33, [[CALC_I_1]] 71 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 72 // ... loop body ... 73 // End of body: store into a[i]: 74 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 75 // CHECK-NOT: !llvm.access.group 76 a[i] = b[i] * c[i] * d[i]; 77 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 78 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 79 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 80 // CHECK-NEXT: br label %{{.+}} 81 } 82 // CHECK: [[LOOP1_END]] 83 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]]) 84 // CHECK-NOT: __kmpc_barrier 85 // CHECK: ret void 86 } 87 88 // CHECK-LABEL: define {{.*void}} @{{.*}}static_not_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 89 void static_not_chunked(float *a, float *b, float *c, float *d) { 90 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 91 #pragma omp for schedule(static) 92 // CHECK: call void @__kmpc_for_static_init_4([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]], i32 34, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 1) 93 // UB = min(UB, GlobalUB) 94 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 95 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423 96 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 97 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 98 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 99 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 100 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 101 // Loop header 102 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 103 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 104 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 105 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 106 for (int i = 32000000; i > 33; i += -7) { 107 // CHECK: [[LOOP1_BODY]] 108 // Start of body: calculate i from IV: 109 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 110 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7 111 // CHECK-NEXT: [[CALC_I_2:%.+]] = sub nsw i32 32000000, [[CALC_I_1]] 112 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 113 // ... loop body ... 114 // End of body: store into a[i]: 115 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 116 // CHECK-NOT: !llvm.access.group 117 a[i] = b[i] * c[i] * d[i]; 118 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 119 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 120 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 121 // CHECK-NEXT: br label %{{.+}} 122 } 123 // CHECK: [[LOOP1_END]] 124 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]]) 125 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 126 // CHECK: ret void 127 } 128 129 // CHECK-LABEL: define {{.*void}} @{{.*}}static_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 130 void static_chunked(float *a, float *b, float *c, float *d) { 131 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 132 #pragma omp for schedule(monotonic: static, 5) 133 // CHECK: call void @__kmpc_for_static_init_4u([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]], i32 536870945, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 5) 134 // UB = min(UB, GlobalUB) 135 // CHECK: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 136 // CHECK-NEXT: [[UBCMP:%.+]] = icmp ugt i32 [[UB]], 16908288 137 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 138 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 16908288, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 139 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 140 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 141 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 142 143 // Outer loop header 144 // CHECK: [[O_IV:%.+]] = load i32, i32* [[OMP_IV]] 145 // CHECK-NEXT: [[O_UB:%.+]] = load i32, i32* [[OMP_UB]] 146 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ule i32 [[O_IV]], [[O_UB]] 147 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 148 149 // Loop header 150 // CHECK: [[O_LOOP1_BODY]] 151 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 152 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 153 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i32 [[IV]], [[UB]] 154 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 155 for (unsigned i = 131071; i <= 2147483647; i += 127) { 156 // CHECK: [[LOOP1_BODY]] 157 // Start of body: calculate i from IV: 158 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 159 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i32 [[IV1_1]], 127 160 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i32 131071, [[CALC_I_1]] 161 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 162 // ... loop body ... 163 // End of body: store into a[i]: 164 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 165 // CHECK-NOT: !llvm.access.group 166 a[i] = b[i] * c[i] * d[i]; 167 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 168 // CHECK-NEXT: [[ADD1_2:%.+]] = add i32 [[IV1_2]], 1 169 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 170 // CHECK-NEXT: br label %{{.+}} 171 } 172 // CHECK: [[LOOP1_END]] 173 // Update the counters, adding stride 174 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 175 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]] 176 // CHECK-NEXT: [[ADD_LB:%.+]] = add i32 [[LB]], [[ST]] 177 // CHECK-NEXT: store i32 [[ADD_LB]], i32* [[OMP_LB]] 178 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 179 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]] 180 // CHECK-NEXT: [[ADD_UB:%.+]] = add i32 [[UB]], [[ST]] 181 // CHECK-NEXT: store i32 [[ADD_UB]], i32* [[OMP_UB]] 182 183 // CHECK: [[O_LOOP1_END]] 184 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]]) 185 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 186 // CHECK: ret void 187 } 188 189 // CHECK-LABEL: define {{.*void}} @{{.*}}dynamic1{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 190 void dynamic1(float *a, float *b, float *c, float *d) { 191 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 192 #pragma omp for schedule(nonmonotonic: dynamic) 193 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741859, i64 0, i64 16908287, i64 1, i64 1) 194 // 195 // CHECK: [[HASWORK:%.+]] = call i32 @__kmpc_dispatch_next_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32* [[OMP_ISLAST:%[^,]+]], i64* [[OMP_LB:%[^,]+]], i64* [[OMP_UB:%[^,]+]], i64* [[OMP_ST:%[^,]+]]) 196 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 197 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 198 199 // Loop header 200 // CHECK: [[O_LOOP1_BODY]] 201 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 202 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 203 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 204 205 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 206 // CHECK-NEXT: [[BOUND:%.+]] = add i64 [[UB]], 1 207 // CHECK-NEXT: [[CMP:%.+]] = icmp ult i64 [[IV]], [[BOUND]] 208 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 209 for (unsigned long long i = 131071; i < 2147483647; i += 127) { 210 // CHECK: [[LOOP1_BODY]] 211 // Start of body: calculate i from IV: 212 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]] 213 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127 214 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]] 215 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]] 216 // ... loop body ... 217 // End of body: store into a[i]: 218 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}!llvm.access.group 219 a[i] = b[i] * c[i] * d[i]; 220 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 221 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1 222 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 223 // CHECK-NEXT: br label %{{.+}} 224 } 225 // CHECK: [[LOOP1_END]] 226 // CHECK: [[O_LOOP1_END]] 227 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 228 // CHECK: ret void 229 } 230 231 // CHECK-LABEL: define {{.*void}} @{{.*}}guided7{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 232 void guided7(float *a, float *b, float *c, float *d) { 233 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 234 #pragma omp for schedule(guided, 7) 235 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 36, i64 0, i64 16908287, i64 1, i64 7) 236 // 237 // CHECK: [[HASWORK:%.+]] = call i32 @__kmpc_dispatch_next_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32* [[OMP_ISLAST:%[^,]+]], i64* [[OMP_LB:%[^,]+]], i64* [[OMP_UB:%[^,]+]], i64* [[OMP_ST:%[^,]+]]) 238 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 239 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 240 241 // Loop header 242 // CHECK: [[O_LOOP1_BODY]] 243 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 244 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 245 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 246 247 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 248 // CHECK-NEXT: [[BOUND:%.+]] = add i64 [[UB]], 1 249 // CHECK-NEXT: [[CMP:%.+]] = icmp ult i64 [[IV]], [[BOUND]] 250 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 251 for (unsigned long long i = 131071; i < 2147483647; i += 127) { 252 // CHECK: [[LOOP1_BODY]] 253 // Start of body: calculate i from IV: 254 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]] 255 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127 256 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]] 257 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]] 258 // ... loop body ... 259 // End of body: store into a[i]: 260 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}!llvm.access.group 261 a[i] = b[i] * c[i] * d[i]; 262 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 263 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1 264 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 265 // CHECK-NEXT: br label %{{.+}} 266 } 267 // CHECK: [[LOOP1_END]] 268 // CHECK: [[O_LOOP1_END]] 269 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 270 // CHECK: ret void 271 } 272 273 // CHECK-LABEL: define {{.*void}} @{{.*}}test_auto{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 274 void test_auto(float *a, float *b, float *c, float *d) { 275 unsigned int x = 0; 276 unsigned int y = 0; 277 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 278 #pragma omp for schedule(auto) collapse(2) 279 // CHECK: call void @__kmpc_dispatch_init_8([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 38, i64 0, i64 [[LAST_ITER:%[^,]+]], i64 1, i64 1) 280 // 281 // CHECK: [[HASWORK:%.+]] = call i32 @__kmpc_dispatch_next_8([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32* [[OMP_ISLAST:%[^,]+]], i64* [[OMP_LB:%[^,]+]], i64* [[OMP_UB:%[^,]+]], i64* [[OMP_ST:%[^,]+]]) 282 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 283 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 284 285 // Loop header 286 // CHECK: [[O_LOOP1_BODY]] 287 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 288 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 289 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 290 291 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 292 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i64 [[IV]], [[UB]] 293 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 294 // FIXME: When the iteration count of some nested loop is not a known constant, 295 // we should pre-calculate it, like we do for the total number of iterations! 296 for (char i = static_cast<char>(y); i <= '9'; ++i) 297 for (x = 11; x > 0; --x) { 298 // CHECK: [[LOOP1_BODY]] 299 // Start of body: indices are calculated from IV: 300 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}} 301 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}} 302 // ... loop body ... 303 // End of body: store into a[i]: 304 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 305 // CHECK-NOT: !llvm.access.group 306 a[i] = b[i] * c[i] * d[i]; 307 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 308 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i64 [[IV1_2]], 1 309 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 310 // CHECK-NEXT: br label %{{.+}} 311 } 312 // CHECK: [[LOOP1_END]] 313 // CHECK: [[O_LOOP1_END]] 314 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 315 // CHECK: ret void 316 } 317 318 // CHECK-LABEL: define {{.*void}} @{{.*}}runtime{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 319 void runtime(float *a, float *b, float *c, float *d) { 320 int x = 0; 321 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 322 #pragma omp for collapse(2) schedule(runtime) 323 // CHECK: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 37, i32 0, i32 199, i32 1, i32 1) 324 // 325 // CHECK: [[HASWORK:%.+]] = call i32 @__kmpc_dispatch_next_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32* [[OMP_ISLAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]]) 326 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 327 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 328 329 // Loop header 330 // CHECK: [[O_LOOP1_BODY]] 331 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 332 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 333 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 334 335 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 336 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 337 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 338 for (unsigned char i = '0' ; i <= '9'; ++i) 339 for (x = -10; x < 10; ++x) { 340 // CHECK: [[LOOP1_BODY]] 341 // Start of body: indices are calculated from IV: 342 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}} 343 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}} 344 // ... loop body ... 345 // End of body: store into a[i]: 346 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 347 // CHECK-NOT: !llvm.access.group 348 a[i] = b[i] * c[i] * d[i]; 349 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 350 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 351 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 352 // CHECK-NEXT: br label %{{.+}} 353 } 354 // CHECK: [[LOOP1_END]] 355 // CHECK: [[O_LOOP1_END]] 356 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 357 // CHECK: ret void 358 } 359 360 // CHECK-LABEL: test_precond 361 void test_precond() { 362 // CHECK: [[A_ADDR:%.+]] = alloca i8, 363 // CHECK: [[I_ADDR:%.+]] = alloca i8, 364 // CHECK: [[CAP:%.+]] = alloca i8, 365 char a = 0; 366 // CHECK: store i8 0, 367 // CHECK: store i32 368 // CHECK: store i8 369 // CHECK: [[A:%.+]] = load i8, i8* [[CAP]], 370 // CHECK: [[CONV:%.+]] = sext i8 [[A]] to i32 371 // CHECK: [[CMP:%.+]] = icmp slt i32 [[CONV]], 10 372 // CHECK: br i1 [[CMP]], label %[[PRECOND_THEN:[^,]+]], label %[[PRECOND_END:[^,]+]] 373 // CHECK: [[PRECOND_THEN]] 374 // CHECK: call void @__kmpc_for_static_init_4 375 #pragma omp for 376 for(char i = a; i < 10; ++i); 377 // CHECK: call void @__kmpc_for_static_fini 378 // CHECK: [[PRECOND_END]] 379 } 380 381 // TERM_DEBUG-LABEL: foo 382 int foo() {return 0;}; 383 384 // TERM_DEBUG-LABEL: parallel_for 385 void parallel_for(float *a) { 386 #pragma omp parallel 387 #pragma omp for schedule(static, 5) 388 // TERM_DEBUG-NOT: __kmpc_global_thread_num 389 // TERM_DEBUG: call void @__kmpc_for_static_init_4u({{.+}}), !dbg [[DBG_LOC:![0-9]+]] 390 // TERM_DEBUG: invoke i32 {{.*}}foo{{.*}}() 391 // TERM_DEBUG: unwind label %[[TERM_LPAD:.+]], 392 // TERM_DEBUG-NOT: __kmpc_global_thread_num 393 // TERM_DEBUG: call void @__kmpc_for_static_fini({{.+}}), !dbg [[DBG_LOC]] 394 // TERM_DEBUG: call {{.+}} @__kmpc_barrier({{.+}}), !dbg [[DBG_LOC]] 395 // TERM_DEBUG: [[TERM_LPAD]] 396 // TERM_DEBUG: call void @__clang_call_terminate 397 // TERM_DEBUG: unreachable 398 for (unsigned i = 131071; i <= 2147483647; i += 127) 399 a[i] += foo(); 400 } 401 // Check source line corresponds to "#pragma omp for schedule(static, 5)" above: 402 // TERM_DEBUG: [[DBG_LOC]] = !DILocation(line: [[@LINE-15]], 403 404 char i = 1, j = 2, k = 3; 405 // CHECK-LABEL: for_with_global_lcv 406 void for_with_global_lcv() { 407 // CHECK: alloca i8, 408 // CHECK: [[I_ADDR:%.+]] = alloca i8, 409 // CHECK: alloca i8, 410 // CHECK: [[J_ADDR:%.+]] = alloca i8, 411 412 // CHECK: call void @__kmpc_for_static_init_4( 413 // CHECK-NOT: [[I]] 414 // CHECK: store i8 %{{.+}}, i8* [[I_ADDR]] 415 // CHECK-NOT: [[I]] 416 // CHECK: [[I_VAL:%.+]] = load i8, i8* [[I_ADDR]], 417 // CHECK-NOT: [[I]] 418 // CHECK: store i8 [[I_VAL]], i8* [[K]] 419 // CHECK-NOT: [[I]] 420 // CHECK: call void @__kmpc_for_static_fini( 421 // CHECK: call void @__kmpc_barrier( 422 #pragma omp for 423 for (i = 0; i < 2; ++i) { 424 k = i; 425 } 426 // CHECK: call void @__kmpc_for_static_init_4( 427 // CHECK-NOT: [[J]] 428 // CHECK: store i8 %{{.+}}, i8* [[J_ADDR]] 429 // CHECK-NOT: [[J]] 430 // CHECK: [[J_VAL:%.+]] = load i8, i8* [[J_ADDR]], 431 // CHECK-NOT: [[J]] 432 // CHECK: store i8 [[J_VAL]], i8* [[K]] 433 // CHECK-NOT: [[J]] 434 // CHECK: call void @__kmpc_for_static_fini( 435 #pragma omp for collapse(2) 436 for (int i = 0; i < 2; ++i) 437 for (j = 0; j < 2; ++j) { 438 k = i; 439 k = j; 440 } 441 char &cnt = i; 442 #pragma omp for 443 for (cnt = 0; cnt < 2; ++cnt) 444 k = cnt; 445 } 446 447 // CHECK-LABEL: for_with_references 448 void for_with_references() { 449 // CHECK: [[I:%.+]] = alloca i8, 450 // CHECK: [[CNT:%.+]] = alloca i8*, 451 // CHECK: [[CNT_PRIV:%.+]] = alloca i8, 452 // CHECK: call void @__kmpc_for_static_init_4( 453 // CHECK-NOT: load i8, i8* [[CNT]], 454 // CHECK: call void @__kmpc_for_static_fini( 455 char i = 0; 456 char &cnt = i; 457 #pragma omp for 458 for (cnt = 0; cnt < 2; ++cnt) 459 k = cnt; 460 } 461 462 struct Bool { 463 Bool(bool b) : b(b) {} 464 operator bool() const { return b; } 465 const bool b; 466 }; 467 468 template <typename T> 469 struct It { 470 It() : p(0) {} 471 It(const It &, int = 0) ; 472 template <typename U> 473 It(U &, int = 0) ; 474 It &operator=(const It &); 475 It &operator=(It &); 476 ~It() {} 477 478 It(T *p) : p(p) {} 479 480 operator T *&() { return p; } 481 operator T *() const { return p; } 482 T *operator->() const { return p; } 483 484 It &operator++() { ++p; return *this; } 485 It &operator--() { --p; return *this; } 486 It &operator+=(unsigned n) { p += n; return *this; } 487 It &operator-=(unsigned n) { p -= n; return *this; } 488 489 T *p; 490 }; 491 492 template <typename T> 493 It<T> operator+(It<T> a, typename It<T>::difference_type n) { return a.p + n; } 494 495 template <typename T> 496 It<T> operator+(typename It<T>::difference_type n, It<T> a) { return a.p + n; } 497 498 template <typename T> 499 It<T> operator-(It<T> a, typename It<T>::difference_type n) { return a.p - n; } 500 501 typedef Bool BoolType; 502 503 template <typename T> 504 BoolType operator<(It<T> a, It<T> b) { return a.p < b.p; } 505 506 void loop_with_It(It<char> begin, It<char> end) { 507 #pragma omp for 508 for (It<char> it = begin; it < end; ++it) { 509 *it = 0; 510 } 511 } 512 513 // CHECK-LABEL: loop_with_It 514 // CHECK: call i32 @__kmpc_global_thread_num( 515 // CHECK: call void @__kmpc_for_static_init_8( 516 // CHECK: call void @__kmpc_for_static_fini( 517 518 void loop_with_It_plus(It<char> begin, It<char> end) { 519 #pragma omp for 520 for (It<char> it = begin; it < end; it+=1u) { 521 *it = 0; 522 } 523 } 524 525 // CHECK-LABEL: loop_with_It_plus 526 // CHECK: call i32 @__kmpc_global_thread_num( 527 // CHECK: call void @__kmpc_for_static_init_8( 528 // CHECK: call void @__kmpc_for_static_fini( 529 530 void loop_with_stmt_expr() { 531 #pragma omp for 532 for (int i = __extension__({float b = 0;b; }); i < __extension__({double c = 1;c; }); i += __extension__({char d = 1; d; })) 533 ; 534 } 535 // CHECK-LABEL: loop_with_stmt_expr 536 // CHECK: call i32 @__kmpc_global_thread_num( 537 // CHECK: call void @__kmpc_for_static_init_4( 538 // CHECK: call void @__kmpc_for_static_fini( 539 540 541 // CHECK-LABEL: fint 542 // CHECK: call {{.*}}i32 {{.*}}ftemplate 543 // CHECK: ret i32 544 545 // CHECK: load i16, i16* 546 // CHECK: store i16 % 547 // CHECK: call void {{.+}}@__kmpc_fork_call( 548 // CHECK: call void @__kmpc_for_static_init_4( 549 template <typename T> 550 T ftemplate() { 551 short aa = 0; 552 553 #pragma omp parallel for schedule(static, aa) 554 for (int i = 0; i < 100; i++) { 555 } 556 return T(); 557 } 558 559 int fint(void) { return ftemplate<int>(); } 560 561 #endif // HEADER 562