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 -debug-info-kind=line-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 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 -verify -triple x86_64-apple-darwin10 -O1 -fopenmp-simd -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s 12 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}} 13 // expected-no-diagnostics 14 #ifndef HEADER 15 #define HEADER 16 17 // CHECK-DAG: [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, i8* } 18 // CHECK-DAG: [[LOOP_LOC:@.+]] = private unnamed_addr global [[IDENT_T_TY]] { i32 0, i32 514, i32 0, i32 0, i8* 19 20 // CHECK-LABEL: with_var_schedule 21 void with_var_schedule() { 22 double a = 5; 23 // CHECK: [[CHUNK_SIZE:%.+]] = fptosi double %{{.+}}to i8 24 // CHECK: store i8 %{{.+}}, i8* [[CHUNK:%.+]], 25 // CHECK: [[VAL:%.+]] = load i8, i8* [[CHUNK]], 26 // CHECK: store i8 [[VAL]], i8* 27 // CHECK: [[CHUNK:%.+]] = load i64, i64* % 28 // CHECK: call void {{.+}} @__kmpc_fork_call({{.+}}, i64 [[CHUNK]]) 29 30 // CHECK: [[CHUNK_VAL:%.+]] = load i8, i8* % 31 // CHECK: [[CHUNK_SIZE:%.+]] = sext i8 [[CHUNK_VAL]] to i64 32 // CHECK: call void @__kmpc_for_static_init_8u([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID:%[^,]+]], i32 33, i32* [[IS_LAST:%[^,]+]], i64* [[OMP_LB:%[^,]+]], i64* [[OMP_UB:%[^,]+]], i64* [[OMP_ST:%[^,]+]], i64 1, i64 [[CHUNK_SIZE]]) 33 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]]) 34 #pragma omp parallel for schedule(static, char(a)) 35 for (unsigned long long i = 1; i < 2; ++i) { 36 } 37 } 38 39 // CHECK-LABEL: define {{.*void}} @{{.*}}without_schedule_clause{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 40 void without_schedule_clause(float *a, float *b, float *c, float *d) { 41 #pragma omp parallel for 42 // 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*, ...)*), 43 // 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) %{{.+}}) 44 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]], 45 // 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) 46 // UB = min(UB, GlobalUB) 47 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 48 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423 49 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 50 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 51 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 52 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 53 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 54 // Loop header 55 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 56 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 57 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 58 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 59 for (int i = 33; i < 32000000; i += 7) { 60 // CHECK: [[LOOP1_BODY]] 61 // Start of body: calculate i from IV: 62 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 63 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7 64 // CHECK-NEXT: [[CALC_I_2:%.+]] = add nsw i32 33, [[CALC_I_1]] 65 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 66 // ... loop body ... 67 // End of body: store into a[i]: 68 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 69 a[i] = b[i] * c[i] * d[i]; 70 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 71 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 72 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 73 // CHECK-NEXT: br label %{{.+}} 74 } 75 // CHECK: [[LOOP1_END]] 76 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]]) 77 // CHECK: ret void 78 } 79 80 // CHECK-LABEL: define {{.*void}} @{{.*}}static_not_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 81 void static_not_chunked(float *a, float *b, float *c, float *d) { 82 #pragma omp parallel for schedule(static) 83 // 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*, ...)*), 84 // 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) %{{.+}}) 85 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]], 86 // 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) 87 // UB = min(UB, GlobalUB) 88 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 89 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423 90 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 91 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 92 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 93 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 94 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 95 // Loop header 96 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 97 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 98 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 99 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 100 for (int i = 32000000; i > 33; i += -7) { 101 // CHECK: [[LOOP1_BODY]] 102 // Start of body: calculate i from IV: 103 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 104 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7 105 // CHECK-NEXT: [[CALC_I_2:%.+]] = sub nsw i32 32000000, [[CALC_I_1]] 106 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 107 // ... loop body ... 108 // End of body: store into a[i]: 109 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 110 a[i] = b[i] * c[i] * d[i]; 111 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 112 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 113 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 114 // CHECK-NEXT: br label %{{.+}} 115 } 116 // CHECK: [[LOOP1_END]] 117 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]]) 118 // CHECK: ret void 119 } 120 121 // CHECK-LABEL: define {{.*void}} @{{.*}}static_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 122 void static_chunked(float *a, float *b, float *c, float *d) { 123 #pragma omp parallel for schedule(static, 5) 124 // 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*, ...)*), 125 // 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) %{{.+}}) 126 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]], 127 // CHECK: call void @__kmpc_for_static_init_4u([[IDENT_T_TY]]* [[LOOP_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]]* [[LOOP_LOC]], i32 [[GTID]]) 178 // CHECK: ret void 179 } 180 181 // CHECK-LABEL: define {{.*void}} @{{.*}}dynamic1{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 182 void dynamic1(float *a, float *b, float *c, float *d) { 183 #pragma omp parallel for schedule(dynamic) 184 // 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*, ...)*), 185 // 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) %{{.+}}) 186 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]], 187 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID:%.+]], i32 35, i64 0, i64 16908287, i64 1, i64 1) 188 // 189 // 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:%[^,]+]]) 190 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 191 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 192 193 // Loop header 194 // CHECK: [[O_LOOP1_BODY]] 195 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 196 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 197 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 198 199 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 200 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i64 [[IV]], [[UB]] 201 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 202 for (unsigned long long i = 131071; i < 2147483647; i += 127) { 203 // CHECK: [[LOOP1_BODY]] 204 // Start of body: calculate i from IV: 205 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]] 206 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127 207 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]] 208 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]] 209 // ... loop body ... 210 // End of body: store into a[i]: 211 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 212 a[i] = b[i] * c[i] * d[i]; 213 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 214 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1 215 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 216 // CHECK-NEXT: br label %{{.+}} 217 } 218 // CHECK: [[LOOP1_END]] 219 // CHECK: [[O_LOOP1_END]] 220 // CHECK: ret void 221 } 222 223 // CHECK-LABEL: define {{.*void}} @{{.*}}guided7{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 224 void guided7(float *a, float *b, float *c, float *d) { 225 #pragma omp parallel for schedule(guided, 7) 226 // 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*, ...)*), 227 // 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) %{{.+}}) 228 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]], 229 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID:%.+]], i32 36, i64 0, i64 16908287, i64 1, i64 7) 230 // 231 // 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:%[^,]+]]) 232 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 233 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 234 235 // Loop header 236 // CHECK: [[O_LOOP1_BODY]] 237 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 238 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 239 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 240 241 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 242 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i64 [[IV]], [[UB]] 243 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 244 for (unsigned long long i = 131071; i < 2147483647; i += 127) { 245 // CHECK: [[LOOP1_BODY]] 246 // Start of body: calculate i from IV: 247 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]] 248 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127 249 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]] 250 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]] 251 // ... loop body ... 252 // End of body: store into a[i]: 253 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 254 a[i] = b[i] * c[i] * d[i]; 255 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 256 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1 257 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 258 // CHECK-NEXT: br label %{{.+}} 259 } 260 // CHECK: [[LOOP1_END]] 261 // CHECK: [[O_LOOP1_END]] 262 // CHECK: ret void 263 } 264 265 // CHECK-LABEL: define {{.*void}} @{{.*}}test_auto{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 266 void test_auto(float *a, float *b, float *c, float *d) { 267 unsigned int x = 0; 268 unsigned int y = 0; 269 #pragma omp parallel for schedule(auto) collapse(2) 270 // 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*, ...)*), 271 // 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) %{{.+}}) 272 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]], 273 // 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) 274 // 275 // 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:%[^,]+]]) 276 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 277 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 278 279 // Loop header 280 // CHECK: [[O_LOOP1_BODY]] 281 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 282 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 283 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 284 285 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 286 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i64 [[IV]], [[UB]] 287 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 288 // FIXME: When the iteration count of some nested loop is not a known constant, 289 // we should pre-calculate it, like we do for the total number of iterations! 290 for (char i = static_cast<char>(y); i <= '9'; ++i) 291 for (x = 11; x > 0; --x) { 292 // CHECK: [[LOOP1_BODY]] 293 // Start of body: indices are calculated from IV: 294 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}} 295 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}} 296 // ... loop body ... 297 // End of body: store into a[i]: 298 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 299 a[i] = b[i] * c[i] * d[i]; 300 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 301 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i64 [[IV1_2]], 1 302 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 303 // CHECK-NEXT: br label %{{.+}} 304 } 305 // CHECK: [[LOOP1_END]] 306 // CHECK: [[O_LOOP1_END]] 307 // CHECK: ret void 308 } 309 310 // CHECK-LABEL: define {{.*void}} @{{.*}}runtime{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 311 void runtime(float *a, float *b, float *c, float *d) { 312 int x = 0; 313 #pragma omp parallel for collapse(2) schedule(runtime) 314 // 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*, ...)*), 315 // 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) %{{.+}}) 316 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]], 317 // CHECK: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID:%.+]], i32 37, i32 0, i32 199, i32 1, i32 1) 318 // 319 // 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:%[^,]+]]) 320 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 321 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 322 323 // Loop header 324 // CHECK: [[O_LOOP1_BODY]] 325 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 326 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 327 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 328 329 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 330 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 331 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 332 for (unsigned char i = '0' ; i <= '9'; ++i) 333 for (x = -10; x < 10; ++x) { 334 // CHECK: [[LOOP1_BODY]] 335 // Start of body: indices are calculated from IV: 336 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}} 337 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}} 338 // ... loop body ... 339 // End of body: store into a[i]: 340 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 341 a[i] = b[i] * c[i] * d[i]; 342 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 343 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 344 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 345 // CHECK-NEXT: br label %{{.+}} 346 } 347 // CHECK: [[LOOP1_END]] 348 // CHECK: [[O_LOOP1_END]] 349 // CHECK: ret void 350 } 351 352 // TERM_DEBUG-LABEL: foo 353 int foo() {return 0;}; 354 355 // TERM_DEBUG-LABEL: parallel_for 356 // CLEANUP: parallel_for 357 void parallel_for(float *a, const int n) { 358 float arr[n]; 359 #pragma omp parallel for schedule(static, 5) private(arr) default(none) firstprivate(n) shared(a) 360 // TERM_DEBUG-NOT: __kmpc_global_thread_num 361 // TERM_DEBUG: call void @__kmpc_for_static_init_4u({{.+}}), !dbg [[DBG_LOC_START:![0-9]+]] 362 // TERM_DEBUG: invoke i32 {{.*}}foo{{.*}}() 363 // TERM_DEBUG: unwind label %[[TERM_LPAD:.+]], 364 // TERM_DEBUG-NOT: __kmpc_global_thread_num 365 // TERM_DEBUG: call void @__kmpc_for_static_fini({{.+}}), !dbg [[DBG_LOC_END:![0-9]+]] 366 // TERM_DEBUG: [[TERM_LPAD]] 367 // TERM_DEBUG: call void @__clang_call_terminate 368 // TERM_DEBUG: unreachable 369 // CLEANUP-NOT: __kmpc_global_thread_num 370 // CLEANUP: call void @__kmpc_for_static_init_4u({{.+}}) 371 // CLEANUP: call void @__kmpc_for_static_fini({{.+}}) 372 for (unsigned i = 131071; i <= 2147483647; i += 127) 373 a[i] += foo() + arr[i] + n; 374 } 375 // Check source line corresponds to "#pragma omp parallel for schedule(static, 5)" above: 376 // TERM_DEBUG-DAG: [[DBG_LOC_START]] = !DILocation(line: [[@LINE-4]], 377 // TERM_DEBUG-DAG: [[DBG_LOC_END]] = !DILocation(line: [[@LINE-18]], 378 379 #endif // HEADER 380 381