1 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s 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 -gline-tables-only -x c++ -emit-llvm %s -o - | FileCheck %s --check-prefix=TERM_DEBUG 5 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -O1 -fopenmp -emit-llvm %s -o - | FileCheck %s --check-prefix=CLEANUP 6 // REQUIRES: x86-registered-target 7 // expected-no-diagnostics 8 #ifndef HEADER 9 #define HEADER 10 11 // CHECK-DAG: [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, i8* } 12 13 // CHECK-LABEL: with_var_schedule 14 void with_var_schedule() { 15 double a = 5; 16 // CHECK: [[CHUNK_SIZE:%.+]] = fptosi double %{{.+}}to i8 17 // CHECK: store i8 %{{.+}}, i8* [[CHUNK:%.+]], 18 // CHECK: call void {{.+}} @__kmpc_fork_call({{.+}}, i8* [[CHUNK]]) 19 20 // CHECK: [[CHUNK:%.+]] = load i8*, i8** % 21 // CHECK: [[CHUNK_VAL:%.+]] = load i8, i8* [[CHUNK]], 22 // CHECK: [[CHUNK_SIZE:%.+]] = sext i8 [[CHUNK_VAL]] to i64 23 // CHECK: call void @__kmpc_for_static_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC:@[^,]+]], i32 [[GTID:%[^,]+]], i32 33, i32* [[IS_LAST:%[^,]+]], i64* [[OMP_LB:%[^,]+]], i64* [[OMP_UB:%[^,]+]], i64* [[OMP_ST:%[^,]+]], i64 1, i64 [[CHUNK_SIZE]]) 24 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]]) 25 // CHECK: __kmpc_barrier 26 #pragma omp parallel for schedule(static, char(a)) 27 for (unsigned long long i = 1; i < 2; ++i) { 28 } 29 } 30 31 // CHECK-LABEL: define {{.*void}} @{{.*}}without_schedule_clause{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 32 void without_schedule_clause(float *a, float *b, float *c, float *d) { 33 #pragma omp parallel for 34 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**, float**, float**, float**)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*), 35 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* noalias [[GTID_PARAM_ADDR:%.+]], i32* noalias %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}) 36 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]], 37 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]], 38 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]], 39 // CHECK: call void @__kmpc_for_static_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 34, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 1) 40 // UB = min(UB, GlobalUB) 41 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 42 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423 43 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 44 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 45 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 46 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 47 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 48 // Loop header 49 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 50 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 51 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 52 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 53 for (int i = 33; i < 32000000; i += 7) { 54 // CHECK: [[LOOP1_BODY]] 55 // Start of body: calculate i from IV: 56 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 57 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7 58 // CHECK-NEXT: [[CALC_I_2:%.+]] = add nsw i32 33, [[CALC_I_1]] 59 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 60 // ... loop body ... 61 // End of body: store into a[i]: 62 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 63 a[i] = b[i] * c[i] * d[i]; 64 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 65 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 66 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 67 // CHECK-NEXT: br label %{{.+}} 68 } 69 // CHECK: [[LOOP1_END]] 70 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]]) 71 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]]) 72 // CHECK: ret void 73 } 74 75 // CHECK-LABEL: define {{.*void}} @{{.*}}static_not_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 76 void static_not_chunked(float *a, float *b, float *c, float *d) { 77 #pragma omp parallel for schedule(static) 78 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**, float**, float**, float**)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*), 79 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* noalias [[GTID_PARAM_ADDR:%.+]], i32* noalias %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}) 80 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]], 81 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]], 82 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]], 83 // CHECK: call void @__kmpc_for_static_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 34, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 1) 84 // UB = min(UB, GlobalUB) 85 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 86 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423 87 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 88 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 89 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 90 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 91 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 92 // Loop header 93 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 94 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 95 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 96 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 97 for (int i = 32000000; i > 33; i += -7) { 98 // CHECK: [[LOOP1_BODY]] 99 // Start of body: calculate i from IV: 100 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 101 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7 102 // CHECK-NEXT: [[CALC_I_2:%.+]] = sub nsw i32 32000000, [[CALC_I_1]] 103 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 104 // ... loop body ... 105 // End of body: store into a[i]: 106 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 107 a[i] = b[i] * c[i] * d[i]; 108 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 109 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 110 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 111 // CHECK-NEXT: br label %{{.+}} 112 } 113 // CHECK: [[LOOP1_END]] 114 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]]) 115 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]]) 116 // CHECK: ret void 117 } 118 119 // CHECK-LABEL: define {{.*void}} @{{.*}}static_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 120 void static_chunked(float *a, float *b, float *c, float *d) { 121 #pragma omp parallel for schedule(static, 5) 122 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**, float**, float**, float**)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*), 123 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* noalias [[GTID_PARAM_ADDR:%.+]], i32* noalias %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}) 124 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]], 125 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]], 126 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]], 127 // CHECK: call void @__kmpc_for_static_init_4u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 33, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 5) 128 // UB = min(UB, GlobalUB) 129 // CHECK: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 130 // CHECK-NEXT: [[UBCMP:%.+]] = icmp ugt i32 [[UB]], 16908288 131 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 132 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 16908288, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 133 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 134 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 135 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 136 137 // Outer loop header 138 // CHECK: [[O_IV:%.+]] = load i32, i32* [[OMP_IV]] 139 // CHECK-NEXT: [[O_UB:%.+]] = load i32, i32* [[OMP_UB]] 140 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ule i32 [[O_IV]], [[O_UB]] 141 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 142 143 // Loop header 144 // CHECK: [[O_LOOP1_BODY]] 145 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 146 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 147 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i32 [[IV]], [[UB]] 148 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 149 for (unsigned i = 131071; i <= 2147483647; i += 127) { 150 // CHECK: [[LOOP1_BODY]] 151 // Start of body: calculate i from IV: 152 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 153 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i32 [[IV1_1]], 127 154 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i32 131071, [[CALC_I_1]] 155 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 156 // ... loop body ... 157 // End of body: store into a[i]: 158 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 159 a[i] = b[i] * c[i] * d[i]; 160 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 161 // CHECK-NEXT: [[ADD1_2:%.+]] = add i32 [[IV1_2]], 1 162 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 163 // CHECK-NEXT: br label %{{.+}} 164 } 165 // CHECK: [[LOOP1_END]] 166 // Update the counters, adding stride 167 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 168 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]] 169 // CHECK-NEXT: [[ADD_LB:%.+]] = add i32 [[LB]], [[ST]] 170 // CHECK-NEXT: store i32 [[ADD_LB]], i32* [[OMP_LB]] 171 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 172 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]] 173 // CHECK-NEXT: [[ADD_UB:%.+]] = add i32 [[UB]], [[ST]] 174 // CHECK-NEXT: store i32 [[ADD_UB]], i32* [[OMP_UB]] 175 176 // CHECK: [[O_LOOP1_END]] 177 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]]) 178 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]]) 179 // CHECK: ret void 180 } 181 182 // CHECK-LABEL: define {{.*void}} @{{.*}}dynamic1{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 183 void dynamic1(float *a, float *b, float *c, float *d) { 184 #pragma omp parallel for schedule(dynamic) 185 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**, float**, float**, float**)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*), 186 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* noalias [[GTID_PARAM_ADDR:%.+]], i32* noalias %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}) 187 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]], 188 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]], 189 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]], 190 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 35, i64 0, i64 16908287, i64 1, i64 1) 191 // 192 // 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:%[^,]+]]) 193 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 194 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 195 196 // Loop header 197 // CHECK: [[O_LOOP1_BODY]] 198 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 199 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 200 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 201 202 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 203 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i64 [[IV]], [[UB]] 204 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 205 for (unsigned long long i = 131071; i < 2147483647; i += 127) { 206 // CHECK: [[LOOP1_BODY]] 207 // Start of body: calculate i from IV: 208 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]] 209 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127 210 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]] 211 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]] 212 // ... loop body ... 213 // End of body: store into a[i]: 214 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 215 a[i] = b[i] * c[i] * d[i]; 216 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 217 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1 218 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 219 // CHECK-NEXT: br label %{{.+}} 220 } 221 // CHECK: [[LOOP1_END]] 222 // CHECK: [[O_LOOP1_END]] 223 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]]) 224 // CHECK: ret void 225 } 226 227 // CHECK-LABEL: define {{.*void}} @{{.*}}guided7{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 228 void guided7(float *a, float *b, float *c, float *d) { 229 #pragma omp parallel for schedule(guided, 7) 230 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**, float**, float**, float**)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*), 231 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* noalias [[GTID_PARAM_ADDR:%.+]], i32* noalias %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}) 232 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]], 233 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]], 234 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]], 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* {{%.+}} 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]]* [[DEFAULT_LOC_BARRIER:[@%].+]], 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 #pragma omp parallel for schedule(auto) collapse(2) 277 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*, float**, float**, float**, float**)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*), 278 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* noalias [[GTID_PARAM_ADDR:%.+]], i32* noalias %{{.+}}, i32* dereferenceable(4) %{{.+}}, i32* dereferenceable(4) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}) 279 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]], 280 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]], 281 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]], 282 // 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) 283 // 284 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]], 285 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]], 286 // 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:%[^,]+]]) 287 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 288 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 289 290 // Loop header 291 // CHECK: [[O_LOOP1_BODY]] 292 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 293 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 294 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 295 296 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 297 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i64 [[IV]], [[UB]] 298 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 299 // FIXME: When the iteration count of some nested loop is not a known constant, 300 // we should pre-calculate it, like we do for the total number of iterations! 301 for (char i = static_cast<char>(y); i <= '9'; ++i) 302 for (x = 11; x > 0; --x) { 303 // CHECK: [[LOOP1_BODY]] 304 // Start of body: indices are calculated from IV: 305 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}} 306 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}} 307 // ... loop body ... 308 // End of body: store into a[i]: 309 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 310 a[i] = b[i] * c[i] * d[i]; 311 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 312 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i64 [[IV1_2]], 1 313 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 314 // CHECK-NEXT: br label %{{.+}} 315 } 316 // CHECK: [[LOOP1_END]] 317 // CHECK: [[O_LOOP1_END]] 318 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]], 319 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]], 320 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]]) 321 // CHECK: ret void 322 } 323 324 // CHECK-LABEL: define {{.*void}} @{{.*}}runtime{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 325 void runtime(float *a, float *b, float *c, float *d) { 326 int x = 0; 327 #pragma omp parallel for collapse(2) schedule(runtime) 328 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, float**, float**, float**, float**)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*), 329 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* noalias [[GTID_PARAM_ADDR:%.+]], i32* noalias %{{.+}}, i32* dereferenceable(4) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}) 330 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]], 331 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]], 332 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]], 333 // CHECK: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 37, i32 0, i32 199, i32 1, i32 1) 334 // 335 // 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:%[^,]+]]) 336 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 337 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 338 339 // Loop header 340 // CHECK: [[O_LOOP1_BODY]] 341 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 342 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 343 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 344 345 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 346 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 347 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 348 for (unsigned char i = '0' ; i <= '9'; ++i) 349 for (x = -10; x < 10; ++x) { 350 // CHECK: [[LOOP1_BODY]] 351 // Start of body: indices are calculated from IV: 352 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}} 353 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}} 354 // ... loop body ... 355 // End of body: store into a[i]: 356 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 357 a[i] = b[i] * c[i] * d[i]; 358 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 359 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 360 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 361 // CHECK-NEXT: br label %{{.+}} 362 } 363 // CHECK: [[LOOP1_END]] 364 // CHECK: [[O_LOOP1_END]] 365 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]], 366 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]], 367 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]]) 368 // CHECK: ret void 369 } 370 371 // TERM_DEBUG-LABEL: foo 372 int foo() {return 0;}; 373 374 // TERM_DEBUG-LABEL: parallel_for 375 // CLEANUP: parallel_for 376 void parallel_for(float *a, int n) { 377 float arr[n]; 378 #pragma omp parallel for schedule(static, 5) private(arr) 379 // TERM_DEBUG-NOT: __kmpc_global_thread_num 380 // TERM_DEBUG: call void @__kmpc_for_static_init_4u({{.+}}), !dbg [[DBG_LOC_START:![0-9]+]] 381 // TERM_DEBUG: invoke i32 {{.*}}foo{{.*}}() 382 // TERM_DEBUG: unwind label %[[TERM_LPAD:.+]], 383 // TERM_DEBUG-NOT: __kmpc_global_thread_num 384 // TERM_DEBUG: call void @__kmpc_for_static_fini({{.+}}), !dbg [[DBG_LOC_END:![0-9]+]] 385 // TERM_DEBUG: call {{.+}} @__kmpc_barrier({{.+}}), !dbg [[DBG_LOC_CANCEL:![0-9]+]] 386 // TERM_DEBUG: [[TERM_LPAD]] 387 // TERM_DEBUG: call void @__clang_call_terminate 388 // TERM_DEBUG: unreachable 389 // CLEANUP-NOT: __kmpc_global_thread_num 390 // CLEANUP: call void @__kmpc_for_static_init_4u({{.+}}) 391 // CLEANUP: call void @__kmpc_for_static_fini({{.+}}) 392 // CLEANUP: call {{.+}} @__kmpc_barrier({{.+}}) 393 for (unsigned i = 131071; i <= 2147483647; i += 127) 394 a[i] += foo() + arr[i]; 395 } 396 // Check source line corresponds to "#pragma omp parallel for schedule(static, 5)" above: 397 // TERM_DEBUG-DAG: [[DBG_LOC_START]] = !DILocation(line: [[@LINE-4]], 398 // TERM_DEBUG-DAG: [[DBG_LOC_END]] = !DILocation(line: [[@LINE-20]], 399 // TERM_DEBUG-DAG: [[DBG_LOC_CANCEL]] = !DILocation(line: [[@LINE-21]], 400 401 #endif // HEADER 402 403