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 -allow-deprecated-dag-overlap %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 -allow-deprecated-dag-overlap %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 -allow-deprecated-dag-overlap %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 -allow-deprecated-dag-overlap %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 -allow-deprecated-dag-overlap --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 -allow-deprecated-dag-overlap --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 -allow-deprecated-dag-overlap --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 -allow-deprecated-dag-overlap --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 // LIFETIME: call void @llvm.lifetime.end 42 #pragma omp for collapse(2) 43 for (int i = 0; i < 4; i++) { 44 for (int j = i; j < 4; j++) { 45 } 46 } 47 } 48 // CHECK-LABEL: define {{.*void}} @{{.*}}without_schedule_clause{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 49 void without_schedule_clause(float *a, float *b, float *c, float *d) { 50 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 51 #pragma omp for nowait 52 // 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) 53 // UB = min(UB, GlobalUB) 54 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 55 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423 56 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 57 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 58 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 59 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 60 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 61 // Loop header 62 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 63 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 64 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 65 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 66 for (int i = 33; i < 32000000; i += 7) { 67 // CHECK: [[LOOP1_BODY]] 68 // Start of body: calculate i from IV: 69 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 70 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7 71 // CHECK-NEXT: [[CALC_I_2:%.+]] = add nsw i32 33, [[CALC_I_1]] 72 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 73 // ... loop body ... 74 // End of body: store into a[i]: 75 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 76 // CHECK-NOT: !llvm.mem.parallel_loop_access 77 a[i] = b[i] * c[i] * d[i]; 78 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 79 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 80 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 81 // CHECK-NEXT: br label %{{.+}} 82 } 83 // CHECK: [[LOOP1_END]] 84 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]]) 85 // CHECK-NOT: __kmpc_barrier 86 // CHECK: ret void 87 } 88 89 // CHECK-LABEL: define {{.*void}} @{{.*}}static_not_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 90 void static_not_chunked(float *a, float *b, float *c, float *d) { 91 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 92 #pragma omp for schedule(static) 93 // 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) 94 // UB = min(UB, GlobalUB) 95 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 96 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423 97 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 98 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 99 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 100 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 101 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 102 // Loop header 103 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 104 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 105 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 106 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 107 for (int i = 32000000; i > 33; i += -7) { 108 // CHECK: [[LOOP1_BODY]] 109 // Start of body: calculate i from IV: 110 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 111 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7 112 // CHECK-NEXT: [[CALC_I_2:%.+]] = sub nsw i32 32000000, [[CALC_I_1]] 113 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 114 // ... loop body ... 115 // End of body: store into a[i]: 116 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 117 // CHECK-NOT: !llvm.mem.parallel_loop_access 118 a[i] = b[i] * c[i] * d[i]; 119 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 120 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 121 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 122 // CHECK-NEXT: br label %{{.+}} 123 } 124 // CHECK: [[LOOP1_END]] 125 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]]) 126 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 127 // CHECK: ret void 128 } 129 130 // CHECK-LABEL: define {{.*void}} @{{.*}}static_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 131 void static_chunked(float *a, float *b, float *c, float *d) { 132 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 133 #pragma omp for schedule(monotonic: static, 5) 134 // 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) 135 // UB = min(UB, GlobalUB) 136 // CHECK: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 137 // CHECK-NEXT: [[UBCMP:%.+]] = icmp ugt i32 [[UB]], 16908288 138 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 139 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 16908288, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 140 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 141 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 142 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 143 144 // Outer loop header 145 // CHECK: [[O_IV:%.+]] = load i32, i32* [[OMP_IV]] 146 // CHECK-NEXT: [[O_UB:%.+]] = load i32, i32* [[OMP_UB]] 147 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ule i32 [[O_IV]], [[O_UB]] 148 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 149 150 // Loop header 151 // CHECK: [[O_LOOP1_BODY]] 152 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 153 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 154 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i32 [[IV]], [[UB]] 155 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 156 for (unsigned i = 131071; i <= 2147483647; i += 127) { 157 // CHECK: [[LOOP1_BODY]] 158 // Start of body: calculate i from IV: 159 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 160 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i32 [[IV1_1]], 127 161 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i32 131071, [[CALC_I_1]] 162 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 163 // ... loop body ... 164 // End of body: store into a[i]: 165 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 166 // CHECK-NOT: !llvm.mem.parallel_loop_access 167 a[i] = b[i] * c[i] * d[i]; 168 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 169 // CHECK-NEXT: [[ADD1_2:%.+]] = add i32 [[IV1_2]], 1 170 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 171 // CHECK-NEXT: br label %{{.+}} 172 } 173 // CHECK: [[LOOP1_END]] 174 // Update the counters, adding stride 175 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 176 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]] 177 // CHECK-NEXT: [[ADD_LB:%.+]] = add i32 [[LB]], [[ST]] 178 // CHECK-NEXT: store i32 [[ADD_LB]], i32* [[OMP_LB]] 179 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 180 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]] 181 // CHECK-NEXT: [[ADD_UB:%.+]] = add i32 [[UB]], [[ST]] 182 // CHECK-NEXT: store i32 [[ADD_UB]], i32* [[OMP_UB]] 183 184 // CHECK: [[O_LOOP1_END]] 185 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]]) 186 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 187 // CHECK: ret void 188 } 189 190 // CHECK-LABEL: define {{.*void}} @{{.*}}dynamic1{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 191 void dynamic1(float *a, float *b, float *c, float *d) { 192 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 193 #pragma omp for schedule(nonmonotonic: dynamic) 194 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741859, i64 0, i64 16908287, i64 1, i64 1) 195 // 196 // 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:%[^,]+]]) 197 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 198 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 199 200 // Loop header 201 // CHECK: [[O_LOOP1_BODY]] 202 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 203 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 204 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 205 206 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 207 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i64 [[IV]], [[UB]] 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.mem.parallel_loop_access 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: [[CMP:%.+]] = icmp ule i64 [[IV]], [[UB]] 249 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 250 for (unsigned long long i = 131071; i < 2147483647; i += 127) { 251 // CHECK: [[LOOP1_BODY]] 252 // Start of body: calculate i from IV: 253 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]] 254 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127 255 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]] 256 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]] 257 // ... loop body ... 258 // End of body: store into a[i]: 259 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}!llvm.mem.parallel_loop_access 260 a[i] = b[i] * c[i] * d[i]; 261 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 262 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1 263 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 264 // CHECK-NEXT: br label %{{.+}} 265 } 266 // CHECK: [[LOOP1_END]] 267 // CHECK: [[O_LOOP1_END]] 268 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 269 // CHECK: ret void 270 } 271 272 // CHECK-LABEL: define {{.*void}} @{{.*}}test_auto{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 273 void test_auto(float *a, float *b, float *c, float *d) { 274 unsigned int x = 0; 275 unsigned int y = 0; 276 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 277 #pragma omp for schedule(auto) collapse(2) 278 // 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) 279 // 280 // 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:%[^,]+]]) 281 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 282 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 283 284 // Loop header 285 // CHECK: [[O_LOOP1_BODY]] 286 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 287 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 288 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 289 290 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 291 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i64 [[IV]], [[UB]] 292 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 293 // FIXME: When the iteration count of some nested loop is not a known constant, 294 // we should pre-calculate it, like we do for the total number of iterations! 295 for (char i = static_cast<char>(y); i <= '9'; ++i) 296 for (x = 11; x > 0; --x) { 297 // CHECK: [[LOOP1_BODY]] 298 // Start of body: indices are calculated from IV: 299 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}} 300 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}} 301 // ... loop body ... 302 // End of body: store into a[i]: 303 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 304 // CHECK-NOT: !llvm.mem.parallel_loop_access 305 a[i] = b[i] * c[i] * d[i]; 306 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 307 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i64 [[IV1_2]], 1 308 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 309 // CHECK-NEXT: br label %{{.+}} 310 } 311 // CHECK: [[LOOP1_END]] 312 // CHECK: [[O_LOOP1_END]] 313 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 314 // CHECK: ret void 315 } 316 317 // CHECK-LABEL: define {{.*void}} @{{.*}}runtime{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 318 void runtime(float *a, float *b, float *c, float *d) { 319 int x = 0; 320 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 321 #pragma omp for collapse(2) schedule(runtime) 322 // CHECK: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 37, i32 0, i32 199, i32 1, i32 1) 323 // 324 // 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:%[^,]+]]) 325 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 326 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 327 328 // Loop header 329 // CHECK: [[O_LOOP1_BODY]] 330 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 331 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 332 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 333 334 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 335 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 336 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 337 for (unsigned char i = '0' ; i <= '9'; ++i) 338 for (x = -10; x < 10; ++x) { 339 // CHECK: [[LOOP1_BODY]] 340 // Start of body: indices are calculated from IV: 341 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}} 342 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}} 343 // ... loop body ... 344 // End of body: store into a[i]: 345 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 346 // CHECK-NOT: !llvm.mem.parallel_loop_access 347 a[i] = b[i] * c[i] * d[i]; 348 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 349 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 350 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 351 // CHECK-NEXT: br label %{{.+}} 352 } 353 // CHECK: [[LOOP1_END]] 354 // CHECK: [[O_LOOP1_END]] 355 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 356 // CHECK: ret void 357 } 358 359 // CHECK-LABEL: test_precond 360 void test_precond() { 361 // CHECK: [[A_ADDR:%.+]] = alloca i8, 362 // CHECK: [[I_ADDR:%.+]] = alloca i8, 363 // CHECK: [[CAP:%.+]] = alloca i8, 364 char a = 0; 365 // CHECK: store i8 0, 366 // CHECK: store i32 367 // CHECK: store i8 368 // CHECK: [[A:%.+]] = load i8, i8* [[CAP]], 369 // CHECK: [[CONV:%.+]] = sext i8 [[A]] to i32 370 // CHECK: [[CMP:%.+]] = icmp slt i32 [[CONV]], 10 371 // CHECK: br i1 [[CMP]], label %[[PRECOND_THEN:[^,]+]], label %[[PRECOND_END:[^,]+]] 372 // CHECK: [[PRECOND_THEN]] 373 // CHECK: call void @__kmpc_for_static_init_4 374 #pragma omp for 375 for(char i = a; i < 10; ++i); 376 // CHECK: call void @__kmpc_for_static_fini 377 // CHECK: [[PRECOND_END]] 378 } 379 380 // TERM_DEBUG-LABEL: foo 381 int foo() {return 0;}; 382 383 // TERM_DEBUG-LABEL: parallel_for 384 void parallel_for(float *a) { 385 #pragma omp parallel 386 #pragma omp for schedule(static, 5) 387 // TERM_DEBUG-NOT: __kmpc_global_thread_num 388 // TERM_DEBUG: call void @__kmpc_for_static_init_4u({{.+}}), !dbg [[DBG_LOC_START:![0-9]+]] 389 // TERM_DEBUG: invoke i32 {{.*}}foo{{.*}}() 390 // TERM_DEBUG: unwind label %[[TERM_LPAD:.+]], 391 // TERM_DEBUG-NOT: __kmpc_global_thread_num 392 // TERM_DEBUG: call void @__kmpc_for_static_fini({{.+}}), !dbg [[DBG_LOC_END:![0-9]+]] 393 // TERM_DEBUG: call {{.+}} @__kmpc_barrier({{.+}}), !dbg [[DBG_LOC_CANCEL:![0-9]+]] 394 // TERM_DEBUG: [[TERM_LPAD]] 395 // TERM_DEBUG: call void @__clang_call_terminate 396 // TERM_DEBUG: unreachable 397 for (unsigned i = 131071; i <= 2147483647; i += 127) 398 a[i] += foo(); 399 } 400 // Check source line corresponds to "#pragma omp for schedule(static, 5)" above: 401 // TERM_DEBUG-DAG: [[DBG_LOC_START]] = !DILocation(line: [[@LINE-15]], 402 // TERM_DEBUG-DAG: [[DBG_LOC_END]] = !DILocation(line: [[@LINE-16]], 403 // TERM_DEBUG-DAG: [[DBG_LOC_CANCEL]] = !DILocation(line: [[@LINE-17]], 404 405 char i = 1, j = 2, k = 3; 406 // CHECK-LABEL: for_with_global_lcv 407 void for_with_global_lcv() { 408 // CHECK: alloca i8, 409 // CHECK: [[I_ADDR:%.+]] = alloca i8, 410 // CHECK: alloca i8, 411 // CHECK: [[J_ADDR:%.+]] = alloca i8, 412 413 // CHECK: call void @__kmpc_for_static_init_4( 414 // CHECK-NOT: [[I]] 415 // CHECK: store i8 %{{.+}}, i8* [[I_ADDR]] 416 // CHECK-NOT: [[I]] 417 // CHECK: [[I_VAL:%.+]] = load i8, i8* [[I_ADDR]], 418 // CHECK-NOT: [[I]] 419 // CHECK: store i8 [[I_VAL]], i8* [[K]] 420 // CHECK-NOT: [[I]] 421 // CHECK: call void @__kmpc_for_static_fini( 422 // CHECK: call void @__kmpc_barrier( 423 #pragma omp for 424 for (i = 0; i < 2; ++i) { 425 k = i; 426 } 427 // CHECK: call void @__kmpc_for_static_init_4( 428 // CHECK-NOT: [[J]] 429 // CHECK: store i8 %{{.+}}, i8* [[J_ADDR]] 430 // CHECK-NOT: [[J]] 431 // CHECK: [[J_VAL:%.+]] = load i8, i8* [[J_ADDR]], 432 // CHECK-NOT: [[J]] 433 // CHECK: store i8 [[J_VAL]], i8* [[K]] 434 // CHECK-NOT: [[J]] 435 // CHECK: call void @__kmpc_for_static_fini( 436 #pragma omp for collapse(2) 437 for (int i = 0; i < 2; ++i) 438 for (j = 0; j < 2; ++j) { 439 k = i; 440 k = j; 441 } 442 char &cnt = i; 443 #pragma omp for 444 for (cnt = 0; cnt < 2; ++cnt) 445 k = cnt; 446 } 447 448 // CHECK-LABEL: for_with_references 449 void for_with_references() { 450 // CHECK: [[I:%.+]] = alloca i8, 451 // CHECK: [[CNT:%.+]] = alloca i8*, 452 // CHECK: [[CNT_PRIV:%.+]] = alloca i8, 453 // CHECK: call void @__kmpc_for_static_init_4( 454 // CHECK-NOT: load i8, i8* [[CNT]], 455 // CHECK: call void @__kmpc_for_static_fini( 456 char i = 0; 457 char &cnt = i; 458 #pragma omp for 459 for (cnt = 0; cnt < 2; ++cnt) 460 k = cnt; 461 } 462 463 struct Bool { 464 Bool(bool b) : b(b) {} 465 operator bool() const { return b; } 466 const bool b; 467 }; 468 469 template <typename T> 470 struct It { 471 It() : p(0) {} 472 It(const It &, int = 0) ; 473 template <typename U> 474 It(U &, int = 0) ; 475 It &operator=(const It &); 476 It &operator=(It &); 477 ~It() {} 478 479 It(T *p) : p(p) {} 480 481 operator T *&() { return p; } 482 operator T *() const { return p; } 483 T *operator->() const { return p; } 484 485 It &operator++() { ++p; return *this; } 486 It &operator--() { --p; return *this; } 487 It &operator+=(unsigned n) { p += n; return *this; } 488 It &operator-=(unsigned n) { p -= n; return *this; } 489 490 T *p; 491 }; 492 493 template <typename T> 494 It<T> operator+(It<T> a, typename It<T>::difference_type n) { return a.p + n; } 495 496 template <typename T> 497 It<T> operator+(typename It<T>::difference_type n, It<T> a) { return a.p + n; } 498 499 template <typename T> 500 It<T> operator-(It<T> a, typename It<T>::difference_type n) { return a.p - n; } 501 502 typedef Bool BoolType; 503 504 template <typename T> 505 BoolType operator<(It<T> a, It<T> b) { return a.p < b.p; } 506 507 void loop_with_It(It<char> begin, It<char> end) { 508 #pragma omp for 509 for (It<char> it = begin; it < end; ++it) { 510 *it = 0; 511 } 512 } 513 514 // CHECK-LABEL: loop_with_It 515 // CHECK: call i32 @__kmpc_global_thread_num( 516 // CHECK: call void @__kmpc_for_static_init_8( 517 // CHECK: call void @__kmpc_for_static_fini( 518 519 void loop_with_It_plus(It<char> begin, It<char> end) { 520 #pragma omp for 521 for (It<char> it = begin; it < end; it+=1u) { 522 *it = 0; 523 } 524 } 525 526 // CHECK-LABEL: loop_with_It_plus 527 // CHECK: call i32 @__kmpc_global_thread_num( 528 // CHECK: call void @__kmpc_for_static_init_8( 529 // CHECK: call void @__kmpc_for_static_fini( 530 531 void loop_with_stmt_expr() { 532 #pragma omp for 533 for (int i = __extension__({float b = 0;b; }); i < __extension__({double c = 1;c; }); i += __extension__({char d = 1; d; })) 534 ; 535 } 536 // CHECK-LABEL: loop_with_stmt_expr 537 // CHECK: call i32 @__kmpc_global_thread_num( 538 // CHECK: call void @__kmpc_for_static_init_4( 539 // CHECK: call void @__kmpc_for_static_fini( 540 541 542 // CHECK-LABEL: fint 543 // CHECK: call {{.*}}i32 {{.*}}ftemplate 544 // CHECK: ret i32 545 546 // CHECK: load i16, i16* 547 // CHECK: store i16 % 548 // CHECK: call void {{.+}}@__kmpc_fork_call( 549 // CHECK: call void @__kmpc_for_static_init_4( 550 template <typename T> 551 T ftemplate() { 552 short aa = 0; 553 554 #pragma omp parallel for schedule(static, aa) 555 for (int i = 0; i < 100; i++) { 556 } 557 return T(); 558 } 559 560 int fint(void) { return ftemplate<int>(); } 561 562 #endif // HEADER 563