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 --check-prefix=OMP45 2 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t -fopenmp-version=50 %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 - -fopenmp-version=50 | FileCheck %s --check-prefix=CHECK --check-prefix=OMP5 4 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s 5 // 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=OMP45 6 // 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 7 // 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 8 9 // 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 10 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t -fopenmp-version=50 %s 11 // 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 - -fopenmp-version=50 | FileCheck --check-prefix SIMD-ONLY0 %s 12 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s 13 // 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 14 // 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 15 // 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 16 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}} 17 // 18 // expected-no-diagnostics 19 #ifndef HEADER 20 #define HEADER 21 // PROF-INSTR-PATH: constant [25 x i8] c"for_codegen-test.profraw\00" 22 23 // CHECK: [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, i8* } 24 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr global %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8* 25 // CHECK-DAG: [[LOOP_LOC:@.+]] = private unnamed_addr global %{{.+}} { i32 0, i32 514, i32 0, i32 0, i8* 26 // CHECK-DAG: [[I:@.+]] = global i8 1, 27 // CHECK-DAG: [[J:@.+]] = global i8 2, 28 // CHECK-DAG: [[K:@.+]] = global i8 3, 29 30 // CHECK-LABEL: loop_with_counter_collapse 31 void loop_with_counter_collapse() { 32 // Captured initializations. 33 // CHECK: store i32 0, i32* [[I_TMP:%.+]], 34 // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]], 35 // CHECK: store i32 [[VAL]], i32* [[J_LB_MIN:%.+]], 36 // CHECK: store i32 3, i32* [[I_TMP]], 37 // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]], 38 // CHECK: store i32 [[VAL]], i32* [[J_LB_MAX:%.+]], 39 // CHECK: [[J_LB_MIN_VAL:%.+]] = load i32, i32* [[J_LB_MIN]], 40 // CHECK: [[J_LB_MAX_VAL:%.+]] = load i32, i32* [[J_LB_MAX]], 41 // CHECK: [[CMP:%.+]] = icmp slt i32 [[J_LB_MIN_VAL]], [[J_LB_MAX_VAL]] 42 // CHECK: [[BOOL:%.+]] = zext i1 [[CMP]] to i8 43 // CHECK: store i8 [[BOOL]], i8* [[J_LB_CMP:%.+]], 44 // CHECK: store i32 0, i32* [[I_TMP]], 45 // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]], 46 // CHECK: [[J_UB_MIN_VAL:%.+]] = add nsw i32 4, [[VAL]] 47 // CHECK: store i32 [[J_UB_MIN_VAL]], i32* [[J_UB_MIN:%.+]], 48 // CHECK: store i32 3, i32* [[I_TMP]], 49 // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]], 50 // CHECK: [[J_UB_MAX_VAL:%.+]] = add nsw i32 4, [[VAL]] 51 // CHECK: store i32 [[J_UB_MAX_VAL]], i32* [[J_UB_MAX:%.+]], 52 // CHECK: [[J_UB_MIN_VAL:%.+]] = load i32, i32* [[J_UB_MIN]], 53 // CHECK: [[J_UB_MAX_VAL:%.+]] = load i32, i32* [[J_UB_MAX]], 54 // CHECK: [[CMP:%.+]] = icmp sgt i32 [[J_UB_MIN_VAL]], [[J_UB_MAX_VAL]] 55 // CHECK: [[BOOL:%.+]] = zext i1 [[CMP]] to i8 56 // CHECK: store i8 [[BOOL]], i8* [[J_UB_CMP:%.+]], 57 // CHECK: [[J_UB_CMP_VAL:%.+]] = load i8, i8* [[J_UB_CMP]], 58 // CHECK: [[BOOL:%.+]] = trunc i8 [[J_UB_CMP_VAL]] to i1 59 // CHECK: br i1 [[BOOL]], label %[[TRUE:[^,]+]], label %[[FALSE:[^,]+]] 60 // CHECK: [[TRUE]]: 61 // CHECK: [[J_UB_MIN_VAL:%.+]] = load i32, i32* [[J_UB_MIN]], 62 // CHECK: br label %[[EXIT:[^,]+]] 63 // CHECK: [[FALSE]]: 64 // CHECK: [[J_UB_MAX_VAL:%.+]] = load i32, i32* [[J_UB_MAX]], 65 // CHECK: br label %[[EXIT]] 66 // CHECK: [[EXIT]]: 67 // CHECK: [[J_UB_VAL:%.+]] = phi i32 [ [[J_UB_MIN_VAL]], %[[TRUE]] ], [ [[J_UB_MAX_VAL]], %[[FALSE]] ] 68 // CHECK: store i32 [[J_UB_VAL]], i32* [[J_UB:%.+]], 69 // CHECK: [[J_LB_CMP_VAL:%.+]] = load i8, i8* [[J_LB_CMP]], 70 // CHECK: [[BOOL:%.+]] = trunc i8 [[J_LB_CMP_VAL]] to i1 71 // CHECK: br i1 [[BOOL]], label %[[TRUE:[^,]+]], label %[[FALSE:[^,]+]] 72 // CHECK: [[TRUE]]: 73 // CHECK: [[J_LB_MIN_VAL:%.+]] = load i32, i32* [[J_LB_MIN]], 74 // CHECK: br label %[[EXIT:[^,]+]] 75 // CHECK: [[FALSE]]: 76 // CHECK: [[J_LB_MAX_VAL:%.+]] = load i32, i32* [[J_LB_MAX]], 77 // CHECK: br label %[[EXIT]] 78 // CHECK: [[EXIT]]: 79 // CHECK: [[J_LB_VAL:%.+]] = phi i32 [ [[J_LB_MIN_VAL]], %[[TRUE]] ], [ [[J_LB_MAX_VAL]], %[[FALSE]] ] 80 // CHECK: store i32 [[J_LB_VAL]], i32* [[J_LB:%.+]], 81 // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]], 82 // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]], 83 // CHECK: [[SUB:%.+]] = sub nsw i32 [[J_UB_VAL]], [[J_LB_VAL]] 84 // CHECK: [[SUB_ST:%.+]] = sub nsw i32 [[SUB]], 1 85 // CHECK: [[ADD_ST:%.+]] = add nsw i32 [[SUB_ST]], 1 86 // CHECK: [[DIV_ST:%.+]] = sdiv i32 [[ADD_ST]], 1 87 // CHECK: [[CAST:%.+]] = sext i32 [[DIV_ST]] to i64 88 // CHECK: [[MUL:%.+]] = mul nsw i64 4, [[CAST]] 89 // CHECK: [[NUM_ITERS_VAL:%.+]] = sub nsw i64 [[MUL]], 1 90 // CHECK: store i64 [[NUM_ITERS_VAL]], i64* [[NUM_ITERS:%.+]], 91 92 // Initialization 93 // CHECK: store i32 0, i32* [[I:%.+]], 94 // CHECK: [[I_INIT:%.+]] = load i32, i32* [[I]], 95 // CHECK: store i32 [[I_INIT]], i32* [[J:%.+]], 96 97 // LIFETIME: call void @llvm.lifetime.end 98 // LIFETIME: call void @llvm.lifetime.end 99 100 // Precondition for j counter 101 // CHECK: store i32 0, i32* [[TMP_I:%.+]], 102 // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[TMP_I]], 103 // CHECK: [[I_VAL:%.+]] = load i32, i32* [[TMP_I]], 104 // CHECK: [[J_UB_VAL:%.+]] = add nsw i32 4, [[I_VAL]] 105 // CHECK: [[CMP:%.+]] = icmp slt i32 [[J_LB_VAL]], [[J_UB_VAL]] 106 // CHECK: br i1 [[CMP]], label %[[THEN:[^,]+]], label %[[ELSE:[^,]+]] 107 108 // CHECK: [[THEN]]: 109 // CHECK: store i64 0, i64* [[LB:%.+]], 110 // CHECK: [[NUM_ITERS_VAL:%.+]] = load i64, i64* [[NUM_ITERS]], 111 // CHECK: store i64 [[NUM_ITERS_VAL]], i64* [[UB:%.+]], 112 // CHECK: store i64 1, i64* [[STRIDE:%.+]], 113 // CHECK: store i32 0, i32* [[IS_LAST:%.+]], 114 // 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) 115 // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]], 116 // CHECK: [[NUM_ITERS_VAL:%.+]] = load i64, i64* [[NUM_ITERS]], 117 // CHECK: [[CMP:%.+]] = icmp sgt i64 [[UB_VAL]], [[NUM_ITERS_VAL]] 118 // CHECK: br i1 [[CMP]], label %[[TRUE:[^,]+]], label %[[FALSE:[^,]+]] 119 // CHECK: [[TRUE]]: 120 // CHECK: [[NUM_ITERS_VAL:%.+]] = load i64, i64* [[NUM_ITERS]], 121 // CHECK: br label %[[DONE:[^,]+]] 122 // CHECK: [[FALSE]]: 123 // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]], 124 // CHECK: br label %[[DONE]] 125 // CHECK: [[DONE]]: 126 // CHECK: [[TOP:%.+]] = phi i64 [ [[NUM_ITERS_VAL]], %[[TRUE]] ], [ [[UB_VAL]], %[[FALSE]] ] 127 // CHECK: store i64 [[TOP]], i64* [[UB]], 128 // CHECK: [[LB_VAL:%.+]] = load i64, i64* [[LB]], 129 // CHECK: store i64 [[LB_VAL]], i64* [[IV:%.+]], 130 // CHECK: br label %[[COND:[^,]+]] 131 // CHECK: [[COND]]: 132 // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]], 133 // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]], 134 // CHECK: [[CMP:%.+]] = icmp sle i64 [[IV_VAL]], [[UB_VAL]] 135 // CHECK: br i1 [[CMP]], label %[[BODY:[^,]+]], label %[[CLEANUP:[^,]+]] 136 // LIFETIME: [[CLEANUP]]: 137 // LIFETIME: br label %[[CLEANUP:[^,]+]] 138 // CHECK: [[BODY]]: 139 // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]], 140 // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]], 141 // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]], 142 // CHECK: [[SUB:%.+]] = sub nsw i32 [[J_UB_VAL]], [[J_LB_VAL]] 143 // CHECK: [[SUB_ST:%.+]] = sub nsw i32 [[SUB]], 1 144 // CHECK: [[ADD_ST:%.+]] = add nsw i32 [[SUB_ST]], 1 145 // CHECK: [[DIV_ST:%.+]] = sdiv i32 [[ADD_ST]], 1 146 // CHECK: [[MUL:%.+]] = mul nsw i32 1, [[DIV_ST]] 147 // CHECK: [[CAST:%.+]] = sext i32 [[MUL]] to i64 148 // CHECK: [[DIV:%.+]] = sdiv i64 [[IV_VAL]], [[CAST]] 149 // CHECK: [[MUL:%.+]] = mul nsw i64 [[DIV]], 1 150 // CHECK: [[ADD:%.+]] = add nsw i64 0, [[MUL]] 151 // CHECK: [[CAST:%.+]] = trunc i64 [[ADD]] to i32 152 // CHECK: store i32 [[CAST]], i32* [[I_PRIV:%.+]], 153 // CHECK: [[I_VAL:%.+]] = load i32, i32* [[I_PRIV]], 154 // CHECK: [[CONV:%.+]] = sext i32 [[I_VAL]] to i64 155 // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]], 156 // CHECK: [[IV_VAL1:%.+]] = load i64, i64* [[IV]], 157 // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]], 158 // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]], 159 // CHECK: [[SUB:%.+]] = sub nsw i32 [[J_UB_VAL]], [[J_LB_VAL]] 160 // CHECK: [[SUB_ST:%.+]] = sub nsw i32 [[SUB]], 1 161 // CHECK: [[ADD_ST:%.+]] = add nsw i32 [[SUB_ST]], 1 162 // CHECK: [[DIV_ST:%.+]] = sdiv i32 [[ADD_ST]], 1 163 // CHECK: [[MUL:%.+]] = mul nsw i32 1, [[DIV_ST]] 164 // CHECK: [[CAST:%.+]] = sext i32 [[MUL]] to i64 165 // CHECK: [[DIV:%.+]] = sdiv i64 [[IV_VAL1]], [[CAST]] 166 // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]], 167 // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]], 168 // CHECK: [[SUB:%.+]] = sub nsw i32 [[J_UB_VAL]], [[J_LB_VAL]] 169 // CHECK: [[SUB_ST:%.+]] = sub nsw i32 [[SUB]], 1 170 // CHECK: [[ADD_ST:%.+]] = add nsw i32 [[SUB_ST]], 1 171 // CHECK: [[DIV_ST:%.+]] = sdiv i32 [[ADD_ST]], 1 172 // CHECK: [[MUL:%.+]] = mul nsw i32 1, [[DIV_ST]] 173 // CHECK: [[CAST:%.+]] = sext i32 [[MUL]] to i64 174 // CHECK: [[MUL:%.+]] = mul nsw i64 [[DIV]], [[CAST]] 175 // CHECK: [[SUB:%.+]] = sub nsw i64 [[IV_VAL]], [[MUL]] 176 // CHECK: [[MUL:%.+]] = mul nsw i64 [[SUB:%.+]], 1 177 // CHECK: [[ADD:%.+]] = add nsw i64 [[CONV]], [[MUL]] 178 // CHECK: [[CAST:%.+]] = trunc i64 [[ADD]] to i32 179 // CHECK: store i32 [[CAST]], i32* [[J_PRIV:%.+]], 180 181 // Check that the loop variable is not out of its boundaries. 182 // CHECK: [[J_VAL:%.+]] = load i32, i32* [[J_PRIV]], 183 // CHECK: [[I_VAL:%.+]] = load i32, i32* [[I_PRIV]], 184 // CHECK: [[J_COND:%.+]] = add nsw i32 4, [[I_VAL]] 185 // CHECK: [[CMP:%.+]] = icmp slt i32 [[J_VAL]], [[J_COND]] 186 // CHECK: br i1 [[CMP]], label %[[NEXT:[^,]+]], label %[[BODY_CONT:[^,]+]] 187 // CHECK: [[NEXT]]: 188 189 // Main body is empty. 190 // CHECK: br label %[[BODY_CONT]] 191 // CHECK: [[BODY_CONT]]: 192 // CHECK: br label %[[INC:[^,]+]] 193 // CHECK: [[INC]]: 194 // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]], 195 // CHECK: [[ADD:%.+]] = add nsw i64 [[IV_VAL]], 1 196 // CHECK: store i64 [[ADD]], i64* [[IV]], 197 // CHECK: br label %[[COND]] 198 // CHECK: [[CLEANUP]]: 199 // CHECK: br label %[[EXIT:[^,]+]] 200 // CHECK: [[EXIT]]: 201 // CHECK: call void @__kmpc_for_static_fini(%struct.ident_t* @{{.+}}, i32 %{{.+}}) 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 // LIFETIME: call void @llvm.lifetime.end 208 // LIFETIME: call void @llvm.lifetime.end 209 // LIFETIME: call void @llvm.lifetime.end 210 // LIFETIME: call void @llvm.lifetime.end 211 #pragma omp for collapse(2) 212 for (int i = 0; i < 4; i++) { 213 for (int j = i; j < 4 + i; j++) { 214 } 215 } 216 } 217 // CHECK-LABEL: define {{.*void}} @{{.*}}without_schedule_clause{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 218 void without_schedule_clause(float *a, float *b, float *c, float *d) { 219 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 220 #pragma omp for nowait 221 // 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) 222 // UB = min(UB, GlobalUB) 223 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 224 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423 225 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 226 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 227 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 228 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 229 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 230 // Loop header 231 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 232 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 233 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 234 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 235 for (int i = 33; i < 32000000; i += 7) { 236 // CHECK: [[LOOP1_BODY]] 237 // Start of body: calculate i from IV: 238 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 239 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7 240 // CHECK-NEXT: [[CALC_I_2:%.+]] = add nsw i32 33, [[CALC_I_1]] 241 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 242 // ... loop body ... 243 // End of body: store into a[i]: 244 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 245 // CHECK-NOT: !llvm.access.group 246 a[i] = b[i] * c[i] * d[i]; 247 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 248 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 249 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 250 // CHECK-NEXT: br label %{{.+}} 251 } 252 // CHECK: [[LOOP1_END]] 253 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]]) 254 // CHECK-NOT: __kmpc_barrier 255 // CHECK: ret void 256 } 257 258 // CHECK-LABEL: define {{.*void}} @{{.*}}static_not_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 259 void static_not_chunked(float *a, float *b, float *c, float *d) { 260 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 261 #pragma omp for schedule(static) 262 // 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) 263 // UB = min(UB, GlobalUB) 264 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 265 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423 266 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 267 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 268 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 269 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 270 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 271 // Loop header 272 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 273 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 274 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 275 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 276 for (int i = 32000000; i > 33; i += -7) { 277 // CHECK: [[LOOP1_BODY]] 278 // Start of body: calculate i from IV: 279 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 280 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7 281 // CHECK-NEXT: [[CALC_I_2:%.+]] = sub nsw i32 32000000, [[CALC_I_1]] 282 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 283 // ... loop body ... 284 // End of body: store into a[i]: 285 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 286 // CHECK-NOT: !llvm.access.group 287 a[i] = b[i] * c[i] * d[i]; 288 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 289 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 290 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 291 // CHECK-NEXT: br label %{{.+}} 292 } 293 // CHECK: [[LOOP1_END]] 294 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]]) 295 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 296 // CHECK: ret void 297 } 298 299 // CHECK-LABEL: define {{.*void}} @{{.*}}static_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 300 void static_chunked(float *a, float *b, float *c, float *d) { 301 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 302 #pragma omp for schedule(monotonic: static, 5) 303 // 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) 304 // UB = min(UB, GlobalUB) 305 // CHECK: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 306 // CHECK-NEXT: [[UBCMP:%.+]] = icmp ugt i32 [[UB]], 16908288 307 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 308 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 16908288, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 309 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 310 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 311 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 312 313 // Outer loop header 314 // CHECK: [[O_IV:%.+]] = load i32, i32* [[OMP_IV]] 315 // CHECK-NEXT: [[O_UB:%.+]] = load i32, i32* [[OMP_UB]] 316 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ule i32 [[O_IV]], [[O_UB]] 317 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 318 319 // Loop header 320 // CHECK: [[O_LOOP1_BODY]] 321 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 322 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 323 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i32 [[IV]], [[UB]] 324 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 325 for (unsigned i = 131071; i <= 2147483647; i += 127) { 326 // CHECK: [[LOOP1_BODY]] 327 // Start of body: calculate i from IV: 328 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]] 329 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i32 [[IV1_1]], 127 330 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i32 131071, [[CALC_I_1]] 331 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 332 // ... loop body ... 333 // End of body: store into a[i]: 334 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 335 // CHECK-NOT: !llvm.access.group 336 a[i] = b[i] * c[i] * d[i]; 337 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 338 // CHECK-NEXT: [[ADD1_2:%.+]] = add i32 [[IV1_2]], 1 339 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 340 // CHECK-NEXT: br label %{{.+}} 341 } 342 // CHECK: [[LOOP1_END]] 343 // Update the counters, adding stride 344 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 345 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]] 346 // CHECK-NEXT: [[ADD_LB:%.+]] = add i32 [[LB]], [[ST]] 347 // CHECK-NEXT: store i32 [[ADD_LB]], i32* [[OMP_LB]] 348 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 349 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]] 350 // CHECK-NEXT: [[ADD_UB:%.+]] = add i32 [[UB]], [[ST]] 351 // CHECK-NEXT: store i32 [[ADD_UB]], i32* [[OMP_UB]] 352 353 // CHECK: [[O_LOOP1_END]] 354 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]]) 355 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 356 // CHECK: ret void 357 } 358 359 // CHECK-LABEL: define {{.*void}} @{{.*}}dynamic1{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 360 void dynamic1(float *a, float *b, float *c, float *d) { 361 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 362 #pragma omp for schedule(nonmonotonic: dynamic) 363 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741859, i64 0, i64 16908287, i64 1, i64 1) 364 // 365 // 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:%[^,]+]]) 366 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 367 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 368 369 // Loop header 370 // CHECK: [[O_LOOP1_BODY]] 371 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 372 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 373 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 374 375 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 376 // CHECK-NEXT: [[BOUND:%.+]] = add i64 [[UB]], 1 377 // CHECK-NEXT: [[CMP:%.+]] = icmp ult i64 [[IV]], [[BOUND]] 378 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 379 for (unsigned long long i = 131071; i < 2147483647; i += 127) { 380 // CHECK: [[LOOP1_BODY]] 381 // Start of body: calculate i from IV: 382 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]] 383 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127 384 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]] 385 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]] 386 // ... loop body ... 387 // End of body: store into a[i]: 388 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}!llvm.access.group 389 a[i] = b[i] * c[i] * d[i]; 390 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 391 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1 392 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 393 // CHECK-NEXT: br label %{{.+}} 394 } 395 // CHECK: [[LOOP1_END]] 396 // CHECK: [[O_LOOP1_END]] 397 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 398 // CHECK: ret void 399 } 400 401 // CHECK-LABEL: define {{.*void}} @{{.*}}guided7{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 402 void guided7(float *a, float *b, float *c, float *d) { 403 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 404 #pragma omp for schedule(guided, 7) 405 // OMP45: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 36, i64 0, i64 16908287, i64 1, i64 7) 406 // OMP5: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741860, i64 0, i64 16908287, i64 1, i64 7) 407 // 408 // 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:%[^,]+]]) 409 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 410 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 411 412 // Loop header 413 // CHECK: [[O_LOOP1_BODY]] 414 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 415 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 416 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 417 418 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 419 // CHECK-NEXT: [[BOUND:%.+]] = add i64 [[UB]], 1 420 // CHECK-NEXT: [[CMP:%.+]] = icmp ult i64 [[IV]], [[BOUND]] 421 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 422 for (unsigned long long i = 131071; i < 2147483647; i += 127) { 423 // CHECK: [[LOOP1_BODY]] 424 // Start of body: calculate i from IV: 425 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]] 426 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127 427 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]] 428 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]] 429 // ... loop body ... 430 // End of body: store into a[i]: 431 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}!llvm.access.group 432 a[i] = b[i] * c[i] * d[i]; 433 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 434 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1 435 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 436 // CHECK-NEXT: br label %{{.+}} 437 } 438 // CHECK: [[LOOP1_END]] 439 // CHECK: [[O_LOOP1_END]] 440 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 441 // CHECK: ret void 442 } 443 444 // CHECK-LABEL: define {{.*void}} @{{.*}}test_auto{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 445 void test_auto(float *a, float *b, float *c, float *d) { 446 unsigned int x = 0; 447 unsigned int y = 0; 448 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 449 #pragma omp for schedule(auto) collapse(2) 450 // 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) 451 // 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) 452 // 453 // 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:%[^,]+]]) 454 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 455 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 456 457 // Loop header 458 // CHECK: [[O_LOOP1_BODY]] 459 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]] 460 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]] 461 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]] 462 463 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]] 464 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i64 [[IV]], [[UB]] 465 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 466 // FIXME: When the iteration count of some nested loop is not a known constant, 467 // we should pre-calculate it, like we do for the total number of iterations! 468 for (char i = static_cast<char>(y); i <= '9'; ++i) 469 for (x = 11; x > 0; --x) { 470 // CHECK: [[LOOP1_BODY]] 471 // Start of body: indices are calculated from IV: 472 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}} 473 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}} 474 // ... loop body ... 475 // End of body: store into a[i]: 476 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 477 // CHECK-NOT: !llvm.access.group 478 a[i] = b[i] * c[i] * d[i]; 479 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}} 480 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i64 [[IV1_2]], 1 481 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]] 482 // CHECK-NEXT: br label %{{.+}} 483 } 484 // CHECK: [[LOOP1_END]] 485 // CHECK: [[O_LOOP1_END]] 486 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 487 // CHECK: ret void 488 } 489 490 // CHECK-LABEL: define {{.*void}} @{{.*}}runtime{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 491 void runtime(float *a, float *b, float *c, float *d) { 492 int x = 0; 493 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 494 #pragma omp for collapse(2) schedule(runtime) 495 // OMP45: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 37, i32 0, i32 199, i32 1, i32 1) 496 // OMP5: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741861, i32 0, i32 199, i32 1, i32 1) 497 // 498 // 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:%[^,]+]]) 499 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0 500 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 501 502 // Loop header 503 // CHECK: [[O_LOOP1_BODY]] 504 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]] 505 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 506 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]] 507 508 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]] 509 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 510 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 511 for (unsigned char i = '0' ; i <= '9'; ++i) 512 for (x = -10; x < 10; ++x) { 513 // CHECK: [[LOOP1_BODY]] 514 // Start of body: indices are calculated from IV: 515 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}} 516 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}} 517 // ... loop body ... 518 // End of body: store into a[i]: 519 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 520 // CHECK-NOT: !llvm.access.group 521 a[i] = b[i] * c[i] * d[i]; 522 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}} 523 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 524 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 525 // CHECK-NEXT: br label %{{.+}} 526 } 527 // CHECK: [[LOOP1_END]] 528 // CHECK: [[O_LOOP1_END]] 529 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]]) 530 // CHECK: ret void 531 } 532 533 // CHECK-LABEL: test_precond 534 void test_precond() { 535 // CHECK: [[A_ADDR:%.+]] = alloca i8, 536 // CHECK: [[I_ADDR:%.+]] = alloca i8, 537 // CHECK: [[CAP:%.+]] = alloca i8, 538 char a = 0; 539 // CHECK: store i8 0, 540 // CHECK: store i32 541 // CHECK: store i8 542 // CHECK: [[A:%.+]] = load i8, i8* [[CAP]], 543 // CHECK: [[CONV:%.+]] = sext i8 [[A]] to i32 544 // CHECK: [[CMP:%.+]] = icmp slt i32 [[CONV]], 10 545 // CHECK: br i1 [[CMP]], label %[[PRECOND_THEN:[^,]+]], label %[[PRECOND_END:[^,]+]] 546 // CHECK: [[PRECOND_THEN]] 547 // CHECK: call void @__kmpc_for_static_init_4 548 #pragma omp for 549 for(char i = a; i < 10; ++i); 550 // CHECK: call void @__kmpc_for_static_fini 551 // CHECK: [[PRECOND_END]] 552 } 553 554 // TERM_DEBUG-LABEL: foo 555 int foo() {return 0;}; 556 557 // TERM_DEBUG-LABEL: parallel_for 558 void parallel_for(float *a) { 559 #pragma omp parallel 560 #pragma omp for schedule(static, 5) 561 // TERM_DEBUG-NOT: __kmpc_global_thread_num 562 // TERM_DEBUG: call void @__kmpc_for_static_init_4u({{.+}}), !dbg [[DBG_LOC:![0-9]+]] 563 // TERM_DEBUG: invoke i32 {{.*}}foo{{.*}}() 564 // TERM_DEBUG: unwind label %[[TERM_LPAD:.+]], 565 // TERM_DEBUG-NOT: __kmpc_global_thread_num 566 // TERM_DEBUG: call void @__kmpc_for_static_fini({{.+}}), !dbg [[DBG_LOC]] 567 // TERM_DEBUG: call {{.+}} @__kmpc_barrier({{.+}}), !dbg [[DBG_LOC]] 568 // TERM_DEBUG: [[TERM_LPAD]] 569 // TERM_DEBUG: call void @__clang_call_terminate 570 // TERM_DEBUG: unreachable 571 for (unsigned i = 131071; i <= 2147483647; i += 127) 572 a[i] += foo(); 573 } 574 // Check source line corresponds to "#pragma omp for schedule(static, 5)" above: 575 // TERM_DEBUG: [[DBG_LOC]] = !DILocation(line: [[@LINE-15]], 576 577 char i = 1, j = 2, k = 3; 578 // CHECK-LABEL: for_with_global_lcv 579 void for_with_global_lcv() { 580 // CHECK: alloca i8, 581 // CHECK: [[I_ADDR:%.+]] = alloca i8, 582 // CHECK: alloca i8, 583 // CHECK: [[J_ADDR:%.+]] = alloca i8, 584 585 // CHECK: call void @__kmpc_for_static_init_4( 586 // CHECK-NOT: [[I]] 587 // CHECK: store i8 %{{.+}}, i8* [[I_ADDR]] 588 // CHECK-NOT: [[I]] 589 // CHECK: [[I_VAL:%.+]] = load i8, i8* [[I_ADDR]], 590 // CHECK-NOT: [[I]] 591 // CHECK: store i8 [[I_VAL]], i8* [[K]] 592 // CHECK-NOT: [[I]] 593 // CHECK: call void @__kmpc_for_static_fini( 594 // CHECK: call void @__kmpc_barrier( 595 #pragma omp for 596 for (i = 0; i < 2; ++i) { 597 k = i; 598 } 599 // CHECK: call void @__kmpc_for_static_init_4( 600 // CHECK-NOT: [[J]] 601 // CHECK: store i8 %{{.+}}, i8* [[J_ADDR]] 602 // CHECK-NOT: [[J]] 603 // CHECK: [[J_VAL:%.+]] = load i8, i8* [[J_ADDR]], 604 // CHECK-NOT: [[J]] 605 // CHECK: store i8 [[J_VAL]], i8* [[K]] 606 // CHECK-NOT: [[J]] 607 // CHECK: call void @__kmpc_for_static_fini( 608 #pragma omp for collapse(2) 609 for (int i = 0; i < 2; ++i) 610 for (j = 0; j < 2; ++j) { 611 k = i; 612 k = j; 613 } 614 char &cnt = i; 615 #pragma omp for 616 for (cnt = 0; cnt < 2; ++cnt) 617 k = cnt; 618 } 619 620 // CHECK-LABEL: for_with_references 621 void for_with_references() { 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 = cnt; 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 #endif // HEADER 737