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