1 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -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 --check-prefix=OMP45 2 // 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 --check-prefix=OMP5 3 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s 4 // 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 --check-prefix=CHECK --check-prefix=OMP5 5 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s 6 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -fopenmp-version=45 -emit-llvm -o - | FileCheck %s --check-prefix=CHECK --check-prefix=OMP45 7 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -fopenmp -fexceptions -fcxx-exceptions -debug-info-kind=line-tables-only -gno-column-info -x c++ -emit-llvm %s -o - | FileCheck %s --check-prefix=TERM_DEBUG 8 // 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 9 10 // 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 11 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s 12 // 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 13 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s 14 // 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 15 // 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 16 // 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 17 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}} 18 // 19 // expected-no-diagnostics 20 #ifndef HEADER 21 #define HEADER 22 // PROF-INSTR-PATH: constant [25 x i8] c"for_codegen-test.profraw\00" 23 24 // CHECK: [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, i8* } 25 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8* 26 // CHECK-DAG: [[LOOP_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 514, i32 0, i32 0, i8* 27 // CHECK-DAG: [[I:@.+]] = global i8 1, 28 // CHECK-DAG: [[J:@.+]] = global i8 2, 29 // CHECK-DAG: [[K:@.+]] = global i8 3, 30 31 // CHECK-LABEL: loop_with_counter_collapse 32 void loop_with_counter_collapse() { 33 // Captured initializations. 34 // CHECK: store i32 0, i32* [[I_TMP:%.+]], 35 // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]], 36 // CHECK: store i32 [[VAL]], i32* [[J_LB_MIN:%.+]], 37 // CHECK: store i32 3, i32* [[I_TMP]], 38 // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]], 39 // CHECK: store i32 [[VAL]], i32* [[J_LB_MAX:%.+]], 40 // CHECK: [[J_LB_MIN_VAL:%.+]] = load i32, i32* [[J_LB_MIN]], 41 // CHECK: [[J_LB_MAX_VAL:%.+]] = load i32, i32* [[J_LB_MAX]], 42 // CHECK: [[CMP:%.+]] = icmp slt i32 [[J_LB_MIN_VAL]], [[J_LB_MAX_VAL]] 43 // CHECK: [[BOOL:%.+]] = zext i1 [[CMP]] to i8 44 // CHECK: store i8 [[BOOL]], i8* [[J_LB_CMP:%.+]], 45 // CHECK: store i32 0, i32* [[I_TMP]], 46 // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]], 47 // CHECK: [[J_UB_MIN_VAL:%.+]] = add nsw i32 4, [[VAL]] 48 // CHECK: store i32 [[J_UB_MIN_VAL]], i32* [[J_UB_MIN:%.+]], 49 // CHECK: store i32 3, i32* [[I_TMP]], 50 // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]], 51 // CHECK: [[J_UB_MAX_VAL:%.+]] = add nsw i32 4, [[VAL]] 52 // CHECK: store i32 [[J_UB_MAX_VAL]], i32* [[J_UB_MAX:%.+]], 53 // CHECK: [[J_UB_MIN_VAL:%.+]] = load i32, i32* [[J_UB_MIN]], 54 // CHECK: [[J_UB_MAX_VAL:%.+]] = load i32, i32* [[J_UB_MAX]], 55 // CHECK: [[CMP:%.+]] = icmp sgt i32 [[J_UB_MIN_VAL]], [[J_UB_MAX_VAL]] 56 // CHECK: [[BOOL:%.+]] = zext i1 [[CMP]] to i8 57 // CHECK: store i8 [[BOOL]], i8* [[J_UB_CMP:%.+]], 58 // CHECK: [[J_UB_CMP_VAL:%.+]] = load i8, i8* [[J_UB_CMP]], 59 // CHECK: [[BOOL:%.+]] = trunc i8 [[J_UB_CMP_VAL]] to i1 60 // CHECK: br i1 [[BOOL]], label %[[TRUE:[^,]+]], label %[[FALSE:[^,]+]] 61 // CHECK: [[TRUE]]: 62 // CHECK: [[J_UB_MIN_VAL:%.+]] = load i32, i32* [[J_UB_MIN]], 63 // CHECK: br label %[[EXIT:[^,]+]] 64 // CHECK: [[FALSE]]: 65 // CHECK: [[J_UB_MAX_VAL:%.+]] = load i32, i32* [[J_UB_MAX]], 66 // CHECK: br label %[[EXIT]] 67 // CHECK: [[EXIT]]: 68 // CHECK: [[J_UB_VAL:%.+]] = phi i32 [ [[J_UB_MIN_VAL]], %[[TRUE]] ], [ [[J_UB_MAX_VAL]], %[[FALSE]] ] 69 // CHECK: store i32 [[J_UB_VAL]], i32* [[J_UB:%.+]], 70 // CHECK: [[J_LB_CMP_VAL:%.+]] = load i8, i8* [[J_LB_CMP]], 71 // CHECK: [[BOOL:%.+]] = trunc i8 [[J_LB_CMP_VAL]] to i1 72 // CHECK: br i1 [[BOOL]], label %[[TRUE:[^,]+]], label %[[FALSE:[^,]+]] 73 // CHECK: [[TRUE]]: 74 // CHECK: [[J_LB_MIN_VAL:%.+]] = load i32, i32* [[J_LB_MIN]], 75 // CHECK: br label %[[EXIT:[^,]+]] 76 // CHECK: [[FALSE]]: 77 // CHECK: [[J_LB_MAX_VAL:%.+]] = load i32, i32* [[J_LB_MAX]], 78 // CHECK: br label %[[EXIT]] 79 // CHECK: [[EXIT]]: 80 // CHECK: [[J_LB_VAL:%.+]] = phi i32 [ [[J_LB_MIN_VAL]], %[[TRUE]] ], [ [[J_LB_MAX_VAL]], %[[FALSE]] ] 81 // CHECK: store i32 [[J_LB_VAL]], i32* [[J_LB:%.+]], 82 // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]], 83 // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]], 84 // CHECK: [[SUB:%.+]] = sub i32 [[J_UB_VAL]], [[J_LB_VAL]] 85 // CHECK: [[SUB_ST:%.+]] = sub i32 [[SUB]], 1 86 // CHECK: [[ADD_ST:%.+]] = add i32 [[SUB_ST]], 1 87 // CHECK: [[DIV_ST:%.+]] = udiv i32 [[ADD_ST]], 1 88 // CHECK: [[CAST:%.+]] = zext i32 [[DIV_ST]] to i64 89 // CHECK: [[MUL:%.+]] = mul nsw i64 4, [[CAST]] 90 // CHECK: [[NUM_ITERS_VAL:%.+]] = sub nsw i64 [[MUL]], 1 91 // CHECK: store i64 [[NUM_ITERS_VAL]], i64* [[NUM_ITERS:%.+]], 92 93 // CHECK: store i64 0, i64* [[LB:%.+]], 94 // CHECK: [[NUM_ITERS_VAL:%.+]] = load i64, i64* [[NUM_ITERS]], 95 // CHECK: store i64 [[NUM_ITERS_VAL]], i64* [[UB:%.+]], 96 // CHECK: store i64 1, i64* [[STRIDE:%.+]], 97 // CHECK: store i32 0, i32* [[IS_LAST:%.+]], 98 // CHECK: call void @__kmpc_for_static_init_8(%struct.ident_t* @{{.+}}, i32 %{{.+}}, i32 34, i32* [[IS_LAST]], i64* [[LB]], i64* [[UB]], i64* [[STRIDE]], i64 1, i64 1) 99 // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]], 100 // CHECK: [[NUM_ITERS_VAL:%.+]] = load i64, i64* [[NUM_ITERS]], 101 // CHECK: [[CMP:%.+]] = icmp sgt i64 [[UB_VAL]], [[NUM_ITERS_VAL]] 102 // CHECK: br i1 [[CMP]], label %[[TRUE:[^,]+]], label %[[FALSE:[^,]+]] 103 // CHECK: [[TRUE]]: 104 // CHECK: [[NUM_ITERS_VAL:%.+]] = load i64, i64* [[NUM_ITERS]], 105 // CHECK: br label %[[DONE:[^,]+]] 106 // CHECK: [[FALSE]]: 107 // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]], 108 // CHECK: br label %[[DONE]] 109 // CHECK: [[DONE]]: 110 // CHECK: [[TOP:%.+]] = phi i64 [ [[NUM_ITERS_VAL]], %[[TRUE]] ], [ [[UB_VAL]], %[[FALSE]] ] 111 // CHECK: store i64 [[TOP]], i64* [[UB]], 112 // CHECK: [[LB_VAL:%.+]] = load i64, i64* [[LB]], 113 // CHECK: store i64 [[LB_VAL]], i64* [[IV:%.+]], 114 // CHECK: br label %[[COND:[^,]+]] 115 // CHECK: [[COND]]: 116 // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]], 117 // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]], 118 // CHECK: [[CMP:%.+]] = icmp sle i64 [[IV_VAL]], [[UB_VAL]] 119 // CHECK: br i1 [[CMP]], label %[[BODY:[^,]+]], label %[[CLEANUP:[^,]+]] 120 // LIFETIME: [[CLEANUP]]: 121 // LIFETIME: br label %[[CLEANUP:[^,]+]] 122 // CHECK: [[BODY]]: 123 // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]], 124 // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]], 125 // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]], 126 // CHECK: [[SUB:%.+]] = sub i32 [[J_UB_VAL]], [[J_LB_VAL]] 127 // CHECK: [[SUB_ST:%.+]] = sub i32 [[SUB]], 1 128 // CHECK: [[ADD_ST:%.+]] = add i32 [[SUB_ST]], 1 129 // CHECK: [[DIV_ST:%.+]] = udiv i32 [[ADD_ST]], 1 130 // CHECK: [[MUL:%.+]] = mul i32 1, [[DIV_ST]] 131 // CHECK: [[CAST:%.+]] = zext i32 [[MUL]] to i64 132 // CHECK: [[DIV:%.+]] = sdiv i64 [[IV_VAL]], [[CAST]] 133 // CHECK: [[MUL:%.+]] = mul nsw i64 [[DIV]], 1 134 // CHECK: [[ADD:%.+]] = add nsw i64 0, [[MUL]] 135 // CHECK: [[CAST:%.+]] = trunc i64 [[ADD]] to i32 136 // CHECK: store i32 [[CAST]], i32* [[I_PRIV:%.+]], 137 // CHECK: [[I_VAL:%.+]] = load i32, i32* [[I_PRIV]], 138 // CHECK: [[CONV:%.+]] = sext i32 [[I_VAL]] to i64 139 // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]], 140 // CHECK: [[IV_VAL1:%.+]] = load i64, i64* [[IV]], 141 // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]], 142 // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]], 143 // CHECK: [[SUB:%.+]] = sub i32 [[J_UB_VAL]], [[J_LB_VAL]] 144 // CHECK: [[SUB_ST:%.+]] = sub i32 [[SUB]], 1 145 // CHECK: [[ADD_ST:%.+]] = add i32 [[SUB_ST]], 1 146 // CHECK: [[DIV_ST:%.+]] = udiv i32 [[ADD_ST]], 1 147 // CHECK: [[MUL:%.+]] = mul i32 1, [[DIV_ST]] 148 // CHECK: [[CAST:%.+]] = zext i32 [[MUL]] to i64 149 // CHECK: [[DIV:%.+]] = sdiv i64 [[IV_VAL1]], [[CAST]] 150 // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]], 151 // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]], 152 // CHECK: [[SUB:%.+]] = sub i32 [[J_UB_VAL]], [[J_LB_VAL]] 153 // CHECK: [[SUB_ST:%.+]] = sub i32 [[SUB]], 1 154 // CHECK: [[ADD_ST:%.+]] = add i32 [[SUB_ST]], 1 155 // CHECK: [[DIV_ST:%.+]] = udiv i32 [[ADD_ST]], 1 156 // CHECK: [[MUL:%.+]] = mul i32 1, [[DIV_ST]] 157 // CHECK: [[CAST:%.+]] = zext i32 [[MUL]] to i64 158 // CHECK: [[MUL:%.+]] = mul nsw i64 [[DIV]], [[CAST]] 159 // CHECK: [[SUB:%.+]] = sub nsw i64 [[IV_VAL]], [[MUL]] 160 // CHECK: [[MUL:%.+]] = mul nsw i64 [[SUB:%.+]], 1 161 // CHECK: [[ADD:%.+]] = add nsw i64 [[CONV]], [[MUL]] 162 // CHECK: [[CAST:%.+]] = trunc i64 [[ADD]] to i32 163 // CHECK: store i32 [[CAST]], i32* [[J_PRIV:%.+]], 164 165 // Check that the loop variable is not out of its boundaries. 166 // CHECK: [[J_VAL:%.+]] = load i32, i32* [[J_PRIV]], 167 // CHECK: [[I_VAL:%.+]] = load i32, i32* [[I_PRIV]], 168 // CHECK: [[J_COND:%.+]] = add nsw i32 4, [[I_VAL]] 169 // CHECK: [[CMP:%.+]] = icmp slt i32 [[J_VAL]], [[J_COND]] 170 // CHECK: br i1 [[CMP]], label %[[NEXT:[^,]+]], label %[[BODY_CONT:[^,]+]] 171 // CHECK: [[NEXT]]: 172 173 // Main body is empty. 174 // CHECK: br label %[[BODY_CONT]] 175 // CHECK: [[BODY_CONT]]: 176 // CHECK: br label %[[INC:[^,]+]] 177 // CHECK: [[INC]]: 178 // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]], 179 // CHECK: [[ADD:%.+]] = add nsw i64 [[IV_VAL]], 1 180 // CHECK: store i64 [[ADD]], i64* [[IV]], 181 // CHECK: br label %[[COND]] 182 // CHECK: [[CLEANUP]]: 183 // CHECK: br label %[[EXIT:[^,]+]] 184 // CHECK: [[EXIT]]: 185 // CHECK: call void @__kmpc_for_static_fini(%struct.ident_t* @{{.+}}, i32 %{{.+}}) 186 // LIFETIME: call void @llvm.lifetime.end 187 // LIFETIME: call void @llvm.lifetime.end 188 // LIFETIME: call void @llvm.lifetime.end 189 // LIFETIME: call void @llvm.lifetime.end 190 // LIFETIME: call void @llvm.lifetime.end 191 // LIFETIME: call void @llvm.lifetime.end 192 // LIFETIME: call void @llvm.lifetime.end 193 // LIFETIME: call void @llvm.lifetime.end 194 // LIFETIME: call void @llvm.lifetime.end 195 #pragma omp for collapse(2) 196 for (int i = 0; i < 4; i++) { 197 for (int j = i; j < 4 + i; j++) { 198 } 199 } 200 } 201 // CHECK-LABEL: define {{.*void}} @{{.*}}without_schedule_clause{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 202 void without_schedule_clause(float *a, float *b, float *c, float *d) { 203 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 204 #pragma omp for nowait 205 // 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) 206 // UB = min(UB, GlobalUB) 207 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 208 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423 209 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 210 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 211 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 212 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 213 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 214 // Loop header 215 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 216 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 217 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 218 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 219 for (int i = 33; i < 32000000; i += 7) { 220 // CHECK: [[LOOP1_BODY]] 221 // Start of body: calculate i from IV: 222 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 223 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7 224 // CHECK-NEXT: [[CALC_I_2:%.+]] = add nsw i32 33, [[CALC_I_1]] 225 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 226 // ... loop body ... 227 // End of body: store into a[i]: 228 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 229 // CHECK-NOT: !llvm.access.group 230 a[i] = b[i] * c[i] * d[i]; 231 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 232 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 233 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 234 // CHECK-NEXT: br label %{{.+}} 235 } 236 // CHECK: [[LOOP1_END]] 237 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]]) 238 // CHECK-NOT: __kmpc_barrier 239 // CHECK: ret void 240 } 241 242 // CHECK-LABEL: define {{.*void}} @{{.*}}static_not_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 243 void static_not_chunked(float *a, float *b, float *c, float *d) { 244 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 245 #pragma omp for schedule(static) 246 // 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) 247 // UB = min(UB, GlobalUB) 248 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 249 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423 250 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 251 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 252 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 253 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 254 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 255 // Loop header 256 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 257 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 258 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 259 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 260 for (int i = 32000000; i > 33; i += -7) { 261 // CHECK: [[LOOP1_BODY]] 262 // Start of body: calculate i from IV: 263 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 264 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7 265 // CHECK-NEXT: [[CALC_I_2:%.+]] = sub nsw i32 32000000, [[CALC_I_1]] 266 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 267 // ... loop body ... 268 // End of body: store into a[i]: 269 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 270 // CHECK-NOT: !llvm.access.group 271 a[i] = b[i] * c[i] * d[i]; 272 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 273 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 274 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 275 // CHECK-NEXT: br label %{{.+}} 276 } 277 // CHECK: [[LOOP1_END]] 278 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]]) 279 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 280 // CHECK: ret void 281 } 282 283 // CHECK-LABEL: define {{.*void}} @{{.*}}static_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 284 void static_chunked(float *a, float *b, float *c, float *d) { 285 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 286 #pragma omp for schedule(monotonic: static, 5) 287 // 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) 288 // UB = min(UB, GlobalUB) 289 // CHECK: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 290 // CHECK-NEXT: [[UBCMP:%.+]] = icmp ugt i32 [[UB]], 16908288 291 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 292 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 16908288, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 293 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 294 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 295 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 296 297 // Outer loop header 298 // CHECK: [[O_IV:%.+]] = load i32, i32* [[OMP_IV]] 299 // CHECK-NEXT: [[O_UB:%.+]] = load i32, i32* [[OMP_UB]] 300 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ule i32 [[O_IV]], [[O_UB]] 301 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 302 303 // Loop header 304 // CHECK: [[O_LOOP1_BODY]] 305 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 306 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 307 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i32 [[IV]], [[UB]] 308 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 309 for (unsigned i = 131071; i <= 2147483647; i += 127) { 310 // CHECK: [[LOOP1_BODY]] 311 // Start of body: calculate i from IV: 312 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 313 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i32 [[IV1_1]], 127 314 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i32 131071, [[CALC_I_1]] 315 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 316 // ... loop body ... 317 // End of body: store into a[i]: 318 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 319 // CHECK-NOT: !llvm.access.group 320 a[i] = b[i] * c[i] * d[i]; 321 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 322 // CHECK-NEXT: [[ADD1_2:%.+]] = add i32 [[IV1_2]], 1 323 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 324 // CHECK-NEXT: br label %{{.+}} 325 } 326 // CHECK: [[LOOP1_END]] 327 // Update the counters, adding stride 328 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 329 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]] 330 // CHECK-NEXT: [[ADD_LB:%.+]] = add i32 [[LB]], [[ST]] 331 // CHECK-NEXT: store i32 [[ADD_LB]], i32* [[OMP_LB]] 332 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 333 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]] 334 // CHECK-NEXT: [[ADD_UB:%.+]] = add i32 [[UB]], [[ST]] 335 // CHECK-NEXT: store i32 [[ADD_UB]], i32* [[OMP_UB]] 336 337 // CHECK: [[O_LOOP1_END]] 338 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]]) 339 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 340 // CHECK: ret void 341 } 342 343 // CHECK-LABEL: define {{.*void}} @{{.*}}dynamic1{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 344 void dynamic1(float *a, float *b, float *c, float *d) { 345 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 346 #pragma omp for schedule(nonmonotonic: dynamic) 347 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741859, i64 0, i64 16908287, i64 1, i64 1) 348 // 349 // 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:%[^,]+]]) 350 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 351 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 352 353 // Loop header 354 // CHECK: [[O_LOOP1_BODY]] 355 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 356 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 357 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 358 359 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 360 // CHECK-NEXT: [[BOUND:%.+]] = add i64 [[UB]], 1 361 // CHECK-NEXT: [[CMP:%.+]] = icmp ult i64 [[IV]], [[BOUND]] 362 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 363 for (unsigned long long i = 131071; i < 2147483647; i += 127) { 364 // CHECK: [[LOOP1_BODY]] 365 // Start of body: calculate i from IV: 366 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]] 367 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127 368 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]] 369 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]] 370 // ... loop body ... 371 // End of body: store into a[i]: 372 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}!llvm.access.group 373 a[i] = b[i] * c[i] * d[i]; 374 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 375 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1 376 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 377 // CHECK-NEXT: br label %{{.+}} 378 } 379 // CHECK: [[LOOP1_END]] 380 // CHECK: [[O_LOOP1_END]] 381 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 382 // CHECK: ret void 383 } 384 385 // CHECK-LABEL: define {{.*void}} @{{.*}}guided7{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 386 void guided7(float *a, float *b, float *c, float *d) { 387 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 388 #pragma omp for schedule(guided, 7) 389 // OMP45: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 36, i64 0, i64 16908287, i64 1, i64 7) 390 // OMP5: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741860, i64 0, i64 16908287, i64 1, i64 7) 391 // 392 // 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:%[^,]+]]) 393 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 394 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 395 396 // Loop header 397 // CHECK: [[O_LOOP1_BODY]] 398 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 399 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 400 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 401 402 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 403 // CHECK-NEXT: [[BOUND:%.+]] = add i64 [[UB]], 1 404 // CHECK-NEXT: [[CMP:%.+]] = icmp ult i64 [[IV]], [[BOUND]] 405 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 406 for (unsigned long long i = 131071; i < 2147483647; i += 127) { 407 // CHECK: [[LOOP1_BODY]] 408 // Start of body: calculate i from IV: 409 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]] 410 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127 411 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]] 412 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]] 413 // ... loop body ... 414 // End of body: store into a[i]: 415 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}!llvm.access.group 416 a[i] = b[i] * c[i] * d[i]; 417 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 418 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1 419 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 420 // CHECK-NEXT: br label %{{.+}} 421 } 422 // CHECK: [[LOOP1_END]] 423 // CHECK: [[O_LOOP1_END]] 424 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 425 // CHECK: ret void 426 } 427 428 // CHECK-LABEL: define {{.*void}} @{{.*}}test_auto{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 429 void test_auto(float *a, float *b, float *c, float *d) { 430 unsigned int x = 0; 431 unsigned int y = 0; 432 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 433 #pragma omp for schedule(auto) collapse(2) 434 // OMP45: call void @__kmpc_dispatch_init_8([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 38, i64 0, i64 [[LAST_ITER:%[^,]+]], i64 1, i64 1) 435 // OMP5: call void @__kmpc_dispatch_init_8([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741862, i64 0, i64 [[LAST_ITER:%[^,]+]], i64 1, i64 1) 436 // 437 // 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:%[^,]+]]) 438 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 439 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 440 441 // Loop header 442 // CHECK: [[O_LOOP1_BODY]] 443 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 444 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 445 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 446 447 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 448 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i64 [[IV]], [[UB]] 449 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 450 // FIXME: When the iteration count of some nested loop is not a known constant, 451 // we should pre-calculate it, like we do for the total number of iterations! 452 for (char i = static_cast<char>(y); i <= '9'; ++i) 453 for (x = 11; x > 0; --x) { 454 // CHECK: [[LOOP1_BODY]] 455 // Start of body: indices are calculated from IV: 456 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}} 457 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}} 458 // ... loop body ... 459 // End of body: store into a[i]: 460 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 461 // CHECK-NOT: !llvm.access.group 462 a[i] = b[i] * c[i] * d[i]; 463 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 464 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i64 [[IV1_2]], 1 465 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 466 // CHECK-NEXT: br label %{{.+}} 467 } 468 // CHECK: [[LOOP1_END]] 469 // CHECK: [[O_LOOP1_END]] 470 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 471 // CHECK: ret void 472 } 473 474 // CHECK-LABEL: define {{.*void}} @{{.*}}runtime{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 475 void runtime(float *a, float *b, float *c, float *d) { 476 int x = 0; 477 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 478 #pragma omp for collapse(2) schedule(runtime) 479 // OMP45: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 37, i32 0, i32 199, i32 1, i32 1) 480 // OMP5: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741861, i32 0, i32 199, i32 1, i32 1) 481 // 482 // 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:%[^,]+]]) 483 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 484 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 485 486 // Loop header 487 // CHECK: [[O_LOOP1_BODY]] 488 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 489 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 490 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 491 492 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 493 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 494 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 495 for (unsigned char i = '0' ; i <= '9'; ++i) 496 for (x = -10; x < 10; ++x) { 497 // CHECK: [[LOOP1_BODY]] 498 // Start of body: indices are calculated from IV: 499 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}} 500 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}} 501 // ... loop body ... 502 // End of body: store into a[i]: 503 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 504 // CHECK-NOT: !llvm.access.group 505 a[i] = b[i] * c[i] * d[i]; 506 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 507 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 508 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 509 // CHECK-NEXT: br label %{{.+}} 510 } 511 // CHECK: [[LOOP1_END]] 512 // CHECK: [[O_LOOP1_END]] 513 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 514 // CHECK: ret void 515 } 516 517 // CHECK-LABEL: test_precond 518 void test_precond() { 519 // CHECK: [[A_ADDR:%.+]] = alloca i8, 520 // CHECK: [[I_ADDR:%.+]] = alloca i8, 521 // CHECK: [[CAP:%.+]] = alloca i8, 522 char a = 0; 523 // CHECK: store i8 0, 524 // CHECK: store i32 525 // CHECK: store i8 526 // CHECK: [[A:%.+]] = load i8, i8* [[CAP]], 527 // CHECK: [[CONV:%.+]] = sext i8 [[A]] to i32 528 // CHECK: [[CMP:%.+]] = icmp slt i32 [[CONV]], 10 529 // CHECK: br i1 [[CMP]], label %[[PRECOND_THEN:[^,]+]], label %[[PRECOND_END:[^,]+]] 530 // CHECK: [[PRECOND_THEN]] 531 // CHECK: call void @__kmpc_for_static_init_4 532 #pragma omp for 533 for(char i = a; i < 10; ++i); 534 // CHECK: call void @__kmpc_for_static_fini 535 // CHECK: [[PRECOND_END]] 536 } 537 538 // TERM_DEBUG-LABEL: foo 539 int foo() { extern void mayThrow(); mayThrow(); return 0;}; 540 541 // TERM_DEBUG-LABEL: parallel_for 542 void parallel_for(float *a) { 543 #pragma omp parallel 544 #pragma omp for schedule(static, 5) 545 // TERM_DEBUG-NOT: __kmpc_global_thread_num 546 // TERM_DEBUG: call void @__kmpc_for_static_init_4u({{.+}}), !dbg [[DBG_LOC:![0-9]+]] 547 // TERM_DEBUG: invoke i32 {{.*}}foo{{.*}}() 548 // TERM_DEBUG: unwind label %[[TERM_LPAD:.+]], 549 // TERM_DEBUG-NOT: __kmpc_global_thread_num 550 // TERM_DEBUG: call void @__kmpc_for_static_fini({{.+}}), !dbg [[DBG_LOC]] 551 // TERM_DEBUG: call {{.+}} @__kmpc_barrier({{.+}}), !dbg [[DBG_LOC]] 552 // TERM_DEBUG: [[TERM_LPAD]] 553 // TERM_DEBUG: call void @__clang_call_terminate 554 // TERM_DEBUG: unreachable 555 for (unsigned i = 131071; i <= 2147483647; i += 127) 556 a[i] += foo(); 557 } 558 // Check source line corresponds to "#pragma omp for schedule(static, 5)" above: 559 // TERM_DEBUG: [[DBG_LOC]] = !DILocation(line: [[@LINE-15]], 560 561 char i = 1, j = 2, k = 3; 562 // CHECK-LABEL: for_with_global_lcv 563 void for_with_global_lcv() { 564 // CHECK: alloca i8, 565 // CHECK: [[I_ADDR:%.+]] = alloca i8, 566 // CHECK: alloca i8, 567 // CHECK: [[J_ADDR:%.+]] = alloca i8, 568 569 // CHECK: call void @__kmpc_for_static_init_4( 570 // CHECK-NOT: [[I]] 571 // CHECK: store i8 %{{.+}}, i8* [[I_ADDR]] 572 // CHECK-NOT: [[I]] 573 // CHECK: [[I_VAL:%.+]] = load i8, i8* [[I_ADDR]], 574 // CHECK-NOT: [[I]] 575 // CHECK: store i8 [[I_VAL]], i8* [[K]] 576 // CHECK-NOT: [[I]] 577 // CHECK: call void @__kmpc_for_static_fini( 578 // CHECK: call void @__kmpc_barrier( 579 #pragma omp for 580 for (i = 0; i < 2; ++i) { 581 k = i; 582 } 583 // CHECK: call void @__kmpc_for_static_init_4( 584 // CHECK-NOT: [[J]] 585 // CHECK: store i8 %{{.+}}, i8* [[J_ADDR]] 586 // CHECK-NOT: [[J]] 587 // CHECK: [[J_VAL:%.+]] = load i8, i8* [[J_ADDR]], 588 // CHECK-NOT: [[J]] 589 // CHECK: store i8 [[J_VAL]], i8* [[K]] 590 // CHECK-NOT: [[J]] 591 // CHECK: call void @__kmpc_for_static_fini( 592 #pragma omp for collapse(2) 593 for (int i = 0; i < 2; ++i) 594 for (j = 0; j < 2; ++j) { 595 k = i; 596 k = j; 597 } 598 char &cnt = i; 599 #pragma omp for 600 for (cnt = 0; cnt < 2; ++cnt) 601 k = cnt; 602 } 603 604 // CHECK-LABEL: for_with_references 605 void for_with_references() { 606 // CHECK: [[I:%.+]] = alloca i8, 607 // CHECK: [[CNT:%.+]] = alloca i8*, 608 // CHECK: [[CNT_PRIV:%.+]] = alloca i8, 609 // CHECK: call void @__kmpc_for_static_init_8( 610 // CHECK-NOT: load i8, i8* [[CNT]], 611 // CHECK: call void @__kmpc_for_static_fini( 612 char i = 0; 613 char &cnt = i; 614 #pragma omp for collapse(2) 615 for (cnt = 0; cnt < 2; ++cnt) 616 for (int j = cnt; j < 4 + cnt; j++) 617 k = cnt; 618 } 619 620 // CHECK-LABEL: for_with_references_dep_cond 621 void for_with_references_dep_cond() { 622 // CHECK: [[I:%.+]] = alloca i8, 623 // CHECK: [[CNT:%.+]] = alloca i8*, 624 // CHECK: [[CNT_PRIV:%.+]] = alloca i8, 625 // CHECK: call void @__kmpc_for_static_init_8( 626 // CHECK-NOT: load i8, i8* [[CNT]], 627 // CHECK: call void @__kmpc_for_static_fini( 628 char i = 0; 629 char &cnt = i; 630 #pragma omp for collapse(2) 631 for (cnt = 0; cnt < 2; ++cnt) 632 for (int j = 0; j < 4 + cnt; j++) 633 k = cnt; 634 } 635 636 struct Bool { 637 Bool(bool b) : b(b) {} 638 operator bool() const { return b; } 639 const bool b; 640 }; 641 642 template <typename T> 643 struct It { 644 It() : p(0) {} 645 It(const It &, int = 0) ; 646 template <typename U> 647 It(U &, int = 0) ; 648 It &operator=(const It &); 649 It &operator=(It &); 650 ~It() {} 651 652 It(T *p) : p(p) {} 653 654 operator T *&() { return p; } 655 operator T *() const { return p; } 656 T *operator->() const { return p; } 657 658 It &operator++() { ++p; return *this; } 659 It &operator--() { --p; return *this; } 660 It &operator+=(unsigned n) { p += n; return *this; } 661 It &operator-=(unsigned n) { p -= n; return *this; } 662 663 T *p; 664 }; 665 666 template <typename T> 667 It<T> operator+(It<T> a, typename It<T>::difference_type n) { return a.p + n; } 668 669 template <typename T> 670 It<T> operator+(typename It<T>::difference_type n, It<T> a) { return a.p + n; } 671 672 template <typename T> 673 It<T> operator-(It<T> a, typename It<T>::difference_type n) { return a.p - n; } 674 675 typedef Bool BoolType; 676 677 template <typename T> 678 BoolType operator<(It<T> a, It<T> b) { return a.p < b.p; } 679 680 void loop_with_It(It<char> begin, It<char> end) { 681 #pragma omp for 682 for (It<char> it = begin; it < end; ++it) { 683 *it = 0; 684 } 685 } 686 687 // CHECK-LABEL: loop_with_It 688 // CHECK: call i32 @__kmpc_global_thread_num( 689 // CHECK: call void @__kmpc_for_static_init_8( 690 // CHECK: call void @__kmpc_for_static_fini( 691 692 void loop_with_It_plus(It<char> begin, It<char> end) { 693 #pragma omp for 694 for (It<char> it = begin; it < end; it+=1u) { 695 *it = 0; 696 } 697 } 698 699 // CHECK-LABEL: loop_with_It_plus 700 // CHECK: call i32 @__kmpc_global_thread_num( 701 // CHECK: call void @__kmpc_for_static_init_8( 702 // CHECK: call void @__kmpc_for_static_fini( 703 704 void loop_with_stmt_expr() { 705 #pragma omp for collapse(2) 706 for (int i = __extension__({float b = 0;b; }); i < __extension__({double c = 1;c; }); i += __extension__({char d = 1; d; })) 707 for (int j = i; j < 4 + i; j++) 708 ; 709 } 710 // CHECK-LABEL: loop_with_stmt_expr 711 // CHECK: call i32 @__kmpc_global_thread_num( 712 // CHECK: call void @__kmpc_for_static_init_8( 713 // CHECK: call void @__kmpc_for_static_fini( 714 715 716 // CHECK-LABEL: fint 717 // CHECK: call {{.*}}i32 {{.*}}ftemplate 718 // CHECK: ret i32 719 720 // CHECK: load i16, i16* 721 // CHECK: store i16 % 722 // CHECK: call void {{.+}}@__kmpc_fork_call( 723 // CHECK: call void @__kmpc_for_static_init_4( 724 template <typename T> 725 T ftemplate() { 726 short aa = 0; 727 728 #pragma omp parallel for schedule(static, aa) 729 for (int i = 0; i < 100; i++) { 730 } 731 return T(); 732 } 733 734 int fint(void) { return ftemplate<int>(); } 735 736 // Check for imperfectly loop nests codegen. 737 #if _OPENMP == 201811 738 void first(); 739 void last(); 740 void inner_f(); 741 void inner_l(); 742 void body_f(); 743 744 // OMP5-LABEL: imperfectly_nested_loop 745 void imperfectly_nested_loop() { 746 // OMP5: call void @__kmpc_for_static_init_4( 747 #pragma omp for collapse(3) order(concurrent) 748 for (int i = 0; i < 10; ++i) { 749 { 750 int a, d; 751 // OMP5: invoke void @{{.+}}first{{.+}}() 752 first(); 753 // OMP5: load i32{{.*}}!llvm.access.group ![[AG:[0-9]+]] 754 // OMP5: store i32{{.*}}!llvm.access.group ![[AG]] 755 a = d; 756 for (int j = 0; j < 10; ++j) { 757 int a, d; 758 // OMP5: invoke void @{{.+}}inner_f{{.+}}() 759 inner_f(); 760 // OMP5: load i32{{.*}}!llvm.access.group ![[AG]] 761 // OMP5: store i32{{.*}}!llvm.access.group ![[AG]] 762 a = d; 763 for (int k = 0; k < 10; ++k) { 764 int a, d; 765 // OMP5: invoke void @{{.+}}body_f{{.+}}() 766 body_f(); 767 // OMP5: load i32{{.*}}!llvm.access.group ![[AG]] 768 // OMP5: store i32{{.*}}!llvm.access.group ![[AG]] 769 a = d; 770 } 771 // OMP5: invoke void @{{.+}}inner_l{{.+}}() 772 inner_l(); 773 } 774 // OMP5: invoke void @{{.+}}last{{.+}}() 775 last(); 776 } 777 } 778 // OMP5: call void @__kmpc_for_static_fini( 779 } 780 781 // OMP5: ![[AG]] = distinct !{} 782 // OMP5: !{!"llvm.loop.parallel_accesses", ![[AG]]} 783 784 #endif 785 786 #endif // HEADER 787