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: [[CMP:%.+]] = icmp ule i64 [[IV]], [[UB]] 207 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 208 for (unsigned long long i = 131071; i < 2147483647; i += 127) { 209 // CHECK: [[LOOP1_BODY]] 210 // Start of body: calculate i from IV: 211 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]] 212 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127 213 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]] 214 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]] 215 // ... loop body ... 216 // End of body: store into a[i]: 217 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}!llvm.access.group 218 a[i] = b[i] * c[i] * d[i]; 219 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 220 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1 221 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 222 // CHECK-NEXT: br label %{{.+}} 223 } 224 // CHECK: [[LOOP1_END]] 225 // CHECK: [[O_LOOP1_END]] 226 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 227 // CHECK: ret void 228 } 229 230 // CHECK-LABEL: define {{.*void}} @{{.*}}guided7{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 231 void guided7(float *a, float *b, float *c, float *d) { 232 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 233 #pragma omp for schedule(guided, 7) 234 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 36, i64 0, i64 16908287, i64 1, i64 7) 235 // 236 // 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:%[^,]+]]) 237 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 238 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 239 240 // Loop header 241 // CHECK: [[O_LOOP1_BODY]] 242 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 243 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 244 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 245 246 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 247 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i64 [[IV]], [[UB]] 248 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 249 for (unsigned long long i = 131071; i < 2147483647; i += 127) { 250 // CHECK: [[LOOP1_BODY]] 251 // Start of body: calculate i from IV: 252 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]] 253 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127 254 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]] 255 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]] 256 // ... loop body ... 257 // End of body: store into a[i]: 258 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}!llvm.access.group 259 a[i] = b[i] * c[i] * d[i]; 260 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 261 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1 262 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 263 // CHECK-NEXT: br label %{{.+}} 264 } 265 // CHECK: [[LOOP1_END]] 266 // CHECK: [[O_LOOP1_END]] 267 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 268 // CHECK: ret void 269 } 270 271 // CHECK-LABEL: define {{.*void}} @{{.*}}test_auto{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 272 void test_auto(float *a, float *b, float *c, float *d) { 273 unsigned int x = 0; 274 unsigned int y = 0; 275 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 276 #pragma omp for schedule(auto) collapse(2) 277 // 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) 278 // 279 // 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:%[^,]+]]) 280 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 281 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 282 283 // Loop header 284 // CHECK: [[O_LOOP1_BODY]] 285 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 286 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 287 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 288 289 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 290 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i64 [[IV]], [[UB]] 291 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 292 // FIXME: When the iteration count of some nested loop is not a known constant, 293 // we should pre-calculate it, like we do for the total number of iterations! 294 for (char i = static_cast<char>(y); i <= '9'; ++i) 295 for (x = 11; x > 0; --x) { 296 // CHECK: [[LOOP1_BODY]] 297 // Start of body: indices are calculated from IV: 298 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}} 299 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}} 300 // ... loop body ... 301 // End of body: store into a[i]: 302 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 303 // CHECK-NOT: !llvm.access.group 304 a[i] = b[i] * c[i] * d[i]; 305 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 306 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i64 [[IV1_2]], 1 307 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 308 // CHECK-NEXT: br label %{{.+}} 309 } 310 // CHECK: [[LOOP1_END]] 311 // CHECK: [[O_LOOP1_END]] 312 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 313 // CHECK: ret void 314 } 315 316 // CHECK-LABEL: define {{.*void}} @{{.*}}runtime{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 317 void runtime(float *a, float *b, float *c, float *d) { 318 int x = 0; 319 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 320 #pragma omp for collapse(2) schedule(runtime) 321 // CHECK: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 37, i32 0, i32 199, i32 1, i32 1) 322 // 323 // 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:%[^,]+]]) 324 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 325 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 326 327 // Loop header 328 // CHECK: [[O_LOOP1_BODY]] 329 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 330 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 331 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 332 333 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 334 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 335 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 336 for (unsigned char i = '0' ; i <= '9'; ++i) 337 for (x = -10; x < 10; ++x) { 338 // CHECK: [[LOOP1_BODY]] 339 // Start of body: indices are calculated from IV: 340 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}} 341 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}} 342 // ... loop body ... 343 // End of body: store into a[i]: 344 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 345 // CHECK-NOT: !llvm.access.group 346 a[i] = b[i] * c[i] * d[i]; 347 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 348 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 349 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 350 // CHECK-NEXT: br label %{{.+}} 351 } 352 // CHECK: [[LOOP1_END]] 353 // CHECK: [[O_LOOP1_END]] 354 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 355 // CHECK: ret void 356 } 357 358 // CHECK-LABEL: test_precond 359 void test_precond() { 360 // CHECK: [[A_ADDR:%.+]] = alloca i8, 361 // CHECK: [[I_ADDR:%.+]] = alloca i8, 362 // CHECK: [[CAP:%.+]] = alloca i8, 363 char a = 0; 364 // CHECK: store i8 0, 365 // CHECK: store i32 366 // CHECK: store i8 367 // CHECK: [[A:%.+]] = load i8, i8* [[CAP]], 368 // CHECK: [[CONV:%.+]] = sext i8 [[A]] to i32 369 // CHECK: [[CMP:%.+]] = icmp slt i32 [[CONV]], 10 370 // CHECK: br i1 [[CMP]], label %[[PRECOND_THEN:[^,]+]], label %[[PRECOND_END:[^,]+]] 371 // CHECK: [[PRECOND_THEN]] 372 // CHECK: call void @__kmpc_for_static_init_4 373 #pragma omp for 374 for(char i = a; i < 10; ++i); 375 // CHECK: call void @__kmpc_for_static_fini 376 // CHECK: [[PRECOND_END]] 377 } 378 379 // TERM_DEBUG-LABEL: foo 380 int foo() {return 0;}; 381 382 // TERM_DEBUG-LABEL: parallel_for 383 void parallel_for(float *a) { 384 #pragma omp parallel 385 #pragma omp for schedule(static, 5) 386 // TERM_DEBUG-NOT: __kmpc_global_thread_num 387 // TERM_DEBUG: call void @__kmpc_for_static_init_4u({{.+}}), !dbg [[DBG_LOC:![0-9]+]] 388 // TERM_DEBUG: invoke i32 {{.*}}foo{{.*}}() 389 // TERM_DEBUG: unwind label %[[TERM_LPAD:.+]], 390 // TERM_DEBUG-NOT: __kmpc_global_thread_num 391 // TERM_DEBUG: call void @__kmpc_for_static_fini({{.+}}), !dbg [[DBG_LOC]] 392 // TERM_DEBUG: call {{.+}} @__kmpc_barrier({{.+}}), !dbg [[DBG_LOC]] 393 // TERM_DEBUG: [[TERM_LPAD]] 394 // TERM_DEBUG: call void @__clang_call_terminate 395 // TERM_DEBUG: unreachable 396 for (unsigned i = 131071; i <= 2147483647; i += 127) 397 a[i] += foo(); 398 } 399 // Check source line corresponds to "#pragma omp for schedule(static, 5)" above: 400 // TERM_DEBUG: [[DBG_LOC]] = !DILocation(line: [[@LINE-15]], 401 402 char i = 1, j = 2, k = 3; 403 // CHECK-LABEL: for_with_global_lcv 404 void for_with_global_lcv() { 405 // CHECK: alloca i8, 406 // CHECK: [[I_ADDR:%.+]] = alloca i8, 407 // CHECK: alloca i8, 408 // CHECK: [[J_ADDR:%.+]] = alloca i8, 409 410 // CHECK: call void @__kmpc_for_static_init_4( 411 // CHECK-NOT: [[I]] 412 // CHECK: store i8 %{{.+}}, i8* [[I_ADDR]] 413 // CHECK-NOT: [[I]] 414 // CHECK: [[I_VAL:%.+]] = load i8, i8* [[I_ADDR]], 415 // CHECK-NOT: [[I]] 416 // CHECK: store i8 [[I_VAL]], i8* [[K]] 417 // CHECK-NOT: [[I]] 418 // CHECK: call void @__kmpc_for_static_fini( 419 // CHECK: call void @__kmpc_barrier( 420 #pragma omp for 421 for (i = 0; i < 2; ++i) { 422 k = i; 423 } 424 // CHECK: call void @__kmpc_for_static_init_4( 425 // CHECK-NOT: [[J]] 426 // CHECK: store i8 %{{.+}}, i8* [[J_ADDR]] 427 // CHECK-NOT: [[J]] 428 // CHECK: [[J_VAL:%.+]] = load i8, i8* [[J_ADDR]], 429 // CHECK-NOT: [[J]] 430 // CHECK: store i8 [[J_VAL]], i8* [[K]] 431 // CHECK-NOT: [[J]] 432 // CHECK: call void @__kmpc_for_static_fini( 433 #pragma omp for collapse(2) 434 for (int i = 0; i < 2; ++i) 435 for (j = 0; j < 2; ++j) { 436 k = i; 437 k = j; 438 } 439 char &cnt = i; 440 #pragma omp for 441 for (cnt = 0; cnt < 2; ++cnt) 442 k = cnt; 443 } 444 445 // CHECK-LABEL: for_with_references 446 void for_with_references() { 447 // CHECK: [[I:%.+]] = alloca i8, 448 // CHECK: [[CNT:%.+]] = alloca i8*, 449 // CHECK: [[CNT_PRIV:%.+]] = alloca i8, 450 // CHECK: call void @__kmpc_for_static_init_4( 451 // CHECK-NOT: load i8, i8* [[CNT]], 452 // CHECK: call void @__kmpc_for_static_fini( 453 char i = 0; 454 char &cnt = i; 455 #pragma omp for 456 for (cnt = 0; cnt < 2; ++cnt) 457 k = cnt; 458 } 459 460 struct Bool { 461 Bool(bool b) : b(b) {} 462 operator bool() const { return b; } 463 const bool b; 464 }; 465 466 template <typename T> 467 struct It { 468 It() : p(0) {} 469 It(const It &, int = 0) ; 470 template <typename U> 471 It(U &, int = 0) ; 472 It &operator=(const It &); 473 It &operator=(It &); 474 ~It() {} 475 476 It(T *p) : p(p) {} 477 478 operator T *&() { return p; } 479 operator T *() const { return p; } 480 T *operator->() const { return p; } 481 482 It &operator++() { ++p; return *this; } 483 It &operator--() { --p; return *this; } 484 It &operator+=(unsigned n) { p += n; return *this; } 485 It &operator-=(unsigned n) { p -= n; return *this; } 486 487 T *p; 488 }; 489 490 template <typename T> 491 It<T> operator+(It<T> a, typename It<T>::difference_type n) { return a.p + n; } 492 493 template <typename T> 494 It<T> operator+(typename It<T>::difference_type n, It<T> a) { return a.p + n; } 495 496 template <typename T> 497 It<T> operator-(It<T> a, typename It<T>::difference_type n) { return a.p - n; } 498 499 typedef Bool BoolType; 500 501 template <typename T> 502 BoolType operator<(It<T> a, It<T> b) { return a.p < b.p; } 503 504 void loop_with_It(It<char> begin, It<char> end) { 505 #pragma omp for 506 for (It<char> it = begin; it < end; ++it) { 507 *it = 0; 508 } 509 } 510 511 // CHECK-LABEL: loop_with_It 512 // CHECK: call i32 @__kmpc_global_thread_num( 513 // CHECK: call void @__kmpc_for_static_init_8( 514 // CHECK: call void @__kmpc_for_static_fini( 515 516 void loop_with_It_plus(It<char> begin, It<char> end) { 517 #pragma omp for 518 for (It<char> it = begin; it < end; it+=1u) { 519 *it = 0; 520 } 521 } 522 523 // CHECK-LABEL: loop_with_It_plus 524 // CHECK: call i32 @__kmpc_global_thread_num( 525 // CHECK: call void @__kmpc_for_static_init_8( 526 // CHECK: call void @__kmpc_for_static_fini( 527 528 void loop_with_stmt_expr() { 529 #pragma omp for 530 for (int i = __extension__({float b = 0;b; }); i < __extension__({double c = 1;c; }); i += __extension__({char d = 1; d; })) 531 ; 532 } 533 // CHECK-LABEL: loop_with_stmt_expr 534 // CHECK: call i32 @__kmpc_global_thread_num( 535 // CHECK: call void @__kmpc_for_static_init_4( 536 // CHECK: call void @__kmpc_for_static_fini( 537 538 539 // CHECK-LABEL: fint 540 // CHECK: call {{.*}}i32 {{.*}}ftemplate 541 // CHECK: ret i32 542 543 // CHECK: load i16, i16* 544 // CHECK: store i16 % 545 // CHECK: call void {{.+}}@__kmpc_fork_call( 546 // CHECK: call void @__kmpc_for_static_init_4( 547 template <typename T> 548 T ftemplate() { 549 short aa = 0; 550 551 #pragma omp parallel for schedule(static, aa) 552 for (int i = 0; i < 100; i++) { 553 } 554 return T(); 555 } 556 557 int fint(void) { return ftemplate<int>(); } 558 559 #endif // HEADER 560